7506R-1C2023-TP2-GRUPO09-CHP01.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "ecf63b83",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package stopwords to /home/fabri/nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    }
   ],
   "source": [
    "try:\n",
    "  import google.colab\n",
    "  IN_COLAB = True\n",
    "except:\n",
    "  IN_COLAB = False\n",
    "import pandas as pd \n",
    "import numpy as np\n",
    "\n",
    "import seaborn as sns\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "from collections import Counter\n",
    "\n",
    "import sklearn\n",
    "from sklearn.model_selection import train_test_split\n",
    "#import sklearn as sk\n",
    "\n",
    "\n",
    "#modelos y metricas\n",
    "import seaborn as sns\n",
    "from matplotlib import pyplot as plt\n",
    "from joblib import dump, load\n",
    "from os.path import exists\n",
    "from sklearn.model_selection import StratifiedKFold, KFold,RandomizedSearchCV, train_test_split, cross_validate\n",
    "from sklearn.metrics import confusion_matrix, classification_report , f1_score, make_scorer, precision_score, recall_score, accuracy_score,f1_score\n",
    "\n",
    "#Xval\n",
    "from sklearn.model_selection import train_test_split, RandomizedSearchCV, GridSearchCV, cross_val_score\n",
    "\n",
    "#vectorizacion\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.ensemble import VotingClassifier\n",
    "from sklearn.pipeline import make_pipeline\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import precision_score, recall_score, accuracy_score,f1_score\n",
    "\n",
    "from joblib import dump, load\n",
    "\n",
    "from os.path import exists\n",
    "from os import environ\n",
    "\n",
    "import string\n",
    "\n",
    "import nltk\n",
    "from sklearn.pipeline import Pipeline\n",
    "\n",
    "import sklearn ### ESTA NO SE BORRA ???? #TODO\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "\n",
    "from os.path import exists\n",
    "environ['TF_CPP_MIN_LOG_LEVEL'] = '3' \n",
    "\n",
    "if IN_COLAB == True:\n",
    "    !pip install nltk\n",
    "\n",
    "import nltk\n",
    "nltk.download('stopwords')\n",
    "stopwords_es = nltk.corpus.stopwords.words('spanish')   \n",
    "    \n",
    "import re\n",
    "from unicodedata import normalize\n",
    "\n",
    "import tensorflow as tfr\n",
    "\n",
    "from tensorflow import keras\n",
    "from keras.preprocessing.text import one_hot, Tokenizer\n",
    "from keras.models import Sequential\n",
    "from keras.layers.core import Activation, Dropout, Dense, SpatialDropout1D\n",
    "from keras.layers import Flatten, GlobalMaxPooling1D, Embedding, Conv1D, LSTM, TextVectorization\n",
    "from keras_preprocessing.sequence import pad_sequences\n",
    "from keras.wrappers.scikit_learn import KerasClassifier\n",
    "import string\n",
    "\n",
    "#Random forest\n",
    "from sklearn.ensemble import RandomForestClassifier "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2ecb45f",
   "metadata": {},
   "source": [
    "Constantes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "9af48c78",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Constantes\n",
    "SEED=9\n",
    "JOBS=-2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "ef7b8dc9",
   "metadata": {},
   "outputs": [],
   "source": [
    "reviewDfOriginal = pd.read_csv(\"./review_train.csv\")\n",
    "reviewDf = reviewDfOriginal.copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "3d101f71",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ID</th>\n",
       "      <th>review_es</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>60000</td>\n",
       "      <td>La mayor virtud de esta película es su existen...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>60001</td>\n",
       "      <td>No soy un experto cinéfilo, pero pocas veces m...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>60002</td>\n",
       "      <td>Si no eres un incondicional del humor estilo T...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>60003</td>\n",
       "      <td>No sé qué está pasando, si la gente se deja ll...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>60004</td>\n",
       "      <td>Pero cuando amanece,y me quedo solo,siento en ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8594</th>\n",
       "      <td>68594</td>\n",
       "      <td>Buena no, lo siguiente. Por fin un film serio ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8595</th>\n",
       "      <td>68595</td>\n",
       "      <td>Me esperaba mucho, pero que mucho, más.Guión m...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8596</th>\n",
       "      <td>68596</td>\n",
       "      <td>De mal cuerpo como sensación al finalizar, de ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8597</th>\n",
       "      <td>68597</td>\n",
       "      <td>Los que han añadido comentarios os lo han dich...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8598</th>\n",
       "      <td>68598</td>\n",
       "      <td>Fui a ver esta película de cine con entusiasmo...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8599 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         ID                                          review_es\n",
       "0     60000  La mayor virtud de esta película es su existen...\n",
       "1     60001  No soy un experto cinéfilo, pero pocas veces m...\n",
       "2     60002  Si no eres un incondicional del humor estilo T...\n",
       "3     60003  No sé qué está pasando, si la gente se deja ll...\n",
       "4     60004  Pero cuando amanece,y me quedo solo,siento en ...\n",
       "...     ...                                                ...\n",
       "8594  68594  Buena no, lo siguiente. Por fin un film serio ...\n",
       "8595  68595  Me esperaba mucho, pero que mucho, más.Guión m...\n",
       "8596  68596  De mal cuerpo como sensación al finalizar, de ...\n",
       "8597  68597  Los que han añadido comentarios os lo han dich...\n",
       "8598  68598  Fui a ver esta película de cine con entusiasmo...\n",
       "\n",
       "[8599 rows x 2 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "review_pruebasOriginal = pd.read_csv(\"./review_test.csv\")\n",
    "review_pruebas = review_pruebasOriginal.copy()\n",
    "review_pruebas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "fc233235",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>review_es</th>\n",
       "      <th>sentimiento</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Uno de los otros críticos ha mencionado que de...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Una pequeña pequeña producción.La técnica de f...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Pensé que esta era una manera maravillosa de p...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Básicamente, hay una familia donde un niño peq...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>El \"amor en el tiempo\" de Petter Mattei es una...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49995</th>\n",
       "      <td>Pensé que esta película hizo un buen trabajo a...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49996</th>\n",
       "      <td>Mala parcela, mal diálogo, mala actuación, dir...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49997</th>\n",
       "      <td>Soy católica enseñada en escuelas primarias pa...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49998</th>\n",
       "      <td>Voy a tener que estar en desacuerdo con el com...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49999</th>\n",
       "      <td>Nadie espera que las películas de Star Trek se...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>50000 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               review_es sentimiento\n",
       "0      Uno de los otros críticos ha mencionado que de...    positivo\n",
       "1      Una pequeña pequeña producción.La técnica de f...    positivo\n",
       "2      Pensé que esta era una manera maravillosa de p...    positivo\n",
       "3      Básicamente, hay una familia donde un niño peq...    negativo\n",
       "4      El \"amor en el tiempo\" de Petter Mattei es una...    positivo\n",
       "...                                                  ...         ...\n",
       "49995  Pensé que esta película hizo un buen trabajo a...    positivo\n",
       "49996  Mala parcela, mal diálogo, mala actuación, dir...    negativo\n",
       "49997  Soy católica enseñada en escuelas primarias pa...    negativo\n",
       "49998  Voy a tener que estar en desacuerdo con el com...    negativo\n",
       "49999  Nadie espera que las películas de Star Trek se...    negativo\n",
       "\n",
       "[50000 rows x 2 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reviewDf = reviewDf.drop([\"ID\"],  axis='columns', inplace=False)\n",
    "reviewDf"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0465afe6",
   "metadata": {},
   "source": [
    "### FILTRO REVIEWS EN INGLES"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "23d75fa4",
   "metadata": {},
   "source": [
    "Filtramos y nos quedamos solo con las reviews en epanol para mejorar la efectividad del modelo.\n",
    "Hay 3 listas hechas a mano debido a que tuvimos que tener cuidado de no incluir palabras cuya raiz exitiera en español, o fueran parte de apellidos ya que en esos casos se veian contenidas (love-lace-, hate-r, excelent-e). Al tener 3 listas con ditintos grupos, fue mas facil armar la lista final. Reconocemos que no es el mejor filtro y que es posible que hayamos eliminado unas pocas reseñas en español y hayan quedado unas pocas en ingles. Sin embargo estamos seguros de que si saca la mayoria de las reseñas en ingles."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "9d81a348",
   "metadata": {
    "lines_to_next_cell": 0
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>review_es</th>\n",
       "      <th>sentimiento</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Uno de los otros críticos ha mencionado que de...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Una pequeña pequeña producción.La técnica de f...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Pensé que esta era una manera maravillosa de p...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Básicamente, hay una familia donde un niño peq...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>El \"amor en el tiempo\" de Petter Mattei es una...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49995</th>\n",
       "      <td>Pensé que esta película hizo un buen trabajo a...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49996</th>\n",
       "      <td>Mala parcela, mal diálogo, mala actuación, dir...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49997</th>\n",
       "      <td>Soy católica enseñada en escuelas primarias pa...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49998</th>\n",
       "      <td>Voy a tener que estar en desacuerdo con el com...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49999</th>\n",
       "      <td>Nadie espera que las películas de Star Trek se...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>48284 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               review_es sentimiento\n",
       "0      Uno de los otros críticos ha mencionado que de...    positivo\n",
       "1      Una pequeña pequeña producción.La técnica de f...    positivo\n",
       "2      Pensé que esta era una manera maravillosa de p...    positivo\n",
       "3      Básicamente, hay una familia donde un niño peq...    negativo\n",
       "4      El \"amor en el tiempo\" de Petter Mattei es una...    positivo\n",
       "...                                                  ...         ...\n",
       "49995  Pensé que esta película hizo un buen trabajo a...    positivo\n",
       "49996  Mala parcela, mal diálogo, mala actuación, dir...    negativo\n",
       "49997  Soy católica enseñada en escuelas primarias pa...    negativo\n",
       "49998  Voy a tener que estar en desacuerdo con el com...    negativo\n",
       "49999  Nadie espera que las películas de Star Trek se...    negativo\n",
       "\n",
       "[48284 rows x 2 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lista_palabras_comunes_en = [\"have\",\"scary\",\"nothing\",\"issue\"]\n",
    "lista_palabras_positivas_en =[\"interesting\",\"amusing\",\"intelligent\",\"pretty\",\"beautiful\"]\n",
    "lista_palabras_negativas_en =[\"boring\",\"bullshit\",\"risk\",\"loss\",\"poor\",\"ugly\"]\n",
    "lista_completa = lista_palabras_comunes_en + lista_palabras_positivas_en + lista_palabras_negativas_en\n",
    "\n",
    "\n",
    "reviewDf_espanol = reviewDf [~reviewDf[\"review_es\"].str.contains('|'.join(lista_completa))]\n",
    "\n",
    "reviewDf_espanol"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "305fc80d",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "b21ebb1a",
   "metadata": {},
   "source": [
    "## Particion"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "37d9668a",
   "metadata": {},
   "source": [
    "Hacemos la particion  en test y train con el dataset ya filtrado"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "b94c1733",
   "metadata": {},
   "outputs": [],
   "source": [
    "reviewDf_x = reviewDf_espanol.drop([\"sentimiento\"],  axis='columns', inplace=False)\n",
    "\n",
    "reviewDf_y = reviewDf_espanol['sentimiento'].copy()\n",
    "\n",
    "x_train, x_test, y_train, y_test = train_test_split(reviewDf_x,\n",
    "                                                    reviewDf_y,\n",
    "                                                    test_size=0.30,\n",
    "                                                    random_state=SEED,\n",
    "                                                    shuffle=True\n",
    "                                                    )"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "db74cc1f",
   "metadata": {},
   "source": [
    "# Bayes Naive\n",
    "\n",
    "Creamos un modelo Bayes Naives con el proposito de observar el comportamiento general del mismo y generamos una primera predicción "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "2ed02e1c",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfTransformer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfVectorizer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator MultinomialNB from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator Pipeline from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "if not exists('modelos/TP2/naiveBayes-n.joblib'):\n",
    "\n",
    "    modeloBayesNaive = make_pipeline(TfidfVectorizer(), MultinomialNB())\n",
    "    modeloBayesNaive.fit(x_train.review_es, y_train)\n",
    "\n",
    "    dump(modeloBayesNaive, 'modelos/TP2/naiveBayes-n.joblib')\n",
    "\n",
    "else:\n",
    "    modeloBayesNaive = load('modelos/TP2/naiveBayes-n.joblib')\n",
    "\n",
    "prediccion = modeloBayesNaive.predict(x_test.review_es)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "82a11817",
   "metadata": {},
   "source": [
    "Analizamos las metricas del modelo para ver cual es el desempeño global del mismo, de manera que, se puede tener una referencia del comportamiento general del modelo a la hora de buscar hiperparametros "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "a2f7bac6",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "    negativo       0.85      0.87      0.86      7300\n",
      "    positivo       0.87      0.84      0.86      7186\n",
      "\n",
      "    accuracy                           0.86     14486\n",
      "   macro avg       0.86      0.86      0.86     14486\n",
      "weighted avg       0.86      0.86      0.86     14486\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(50.722222222222214, 0.5, 'Verdadero')"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#performance\n",
    "print(classification_report(y_test,prediccion))\n",
    "\n",
    "#Creamos la matriz de confusión\n",
    "tablota=confusion_matrix(y_test, prediccion)\n",
    "\n",
    "#Grafico la matriz de confusión\n",
    "sns.heatmap(tablota,cmap='GnBu',annot=True,fmt='g')\n",
    "plt.xlabel('Predicho')\n",
    "plt.ylabel('Verdadero')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9a3bf8c3",
   "metadata": {},
   "source": [
    "Generamos una predicción para Kaggle con nuestro primer modelo, el cual tiene un desempeño aceptable "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "d9067c36",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('submissions/TP2/naiveBayes-nuevo.csv'):\n",
    "    prediccionTesteo = modeloBayesNaive.predict(review_pruebas.review_es)\n",
    "    df_submission = pd.DataFrame({'id': review_pruebasOriginal['ID'], 'sentimiento': prediccionTesteo})\n",
    "    df_submission.to_csv('submissions/TP2/naiveBayes-nuevo.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f98172c9",
   "metadata": {},
   "source": [
    "### Busqueda de hiperparametros \n",
    "\n",
    "Hacemos un pequeño ajuste sobre los datos de entrada en esta ocación "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "ffd93059",
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test_transformed = y_test.map(lambda x: 0 if x == 'positivo' else 1)\n",
    "y_train_transformed = y_train.map(lambda x: 0 if x == 'positivo' else 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f639100d",
   "metadata": {},
   "source": [
    "Para la busqueda de hiperparametros del modelo bayes naives, se debe considerar realizar en simultaneo la optimización tanto del modelo BN como encontrar las condiciones para que la tecnica TF-IDF, implementada por el TfidfVectorizer, mejore las metricas en el analisis de sentimientos\n",
    "\n",
    "Por lo tanto realizamos la busqueda de hiperparametros usando valores de prueba para intentar mejorar el modelo y por otro lado valores de prueba para el preprocesamiento del texto "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "8723dc9e",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "if not exists('modelos/TP2/nb_hiper.joblib'):\n",
    "    frecuencia_maxima = [0.9, 0.95, 0.85, 0.98]\n",
    "    frecuencia_minima = [0.05, 0.07, 0.1, 0.03]\n",
    "    ngram_range = [(1,1), (1,2), (3,1)]\n",
    "    alpha = [0.1, 0.5, 1.0, 2.0, 0.8, 0.3, 1.5]\n",
    "    fit_prior = [True, False]\n",
    "    class_prior = [None, [0.2, 0.8], [0.5, 0.5]]\n",
    "\n",
    "    metrica = make_scorer(sklearn.metrics.f1_score)\n",
    "    folds = 5\n",
    "\n",
    "    modelo = Pipeline(steps=[('tfidfvectorizer', TfidfVectorizer()), ('multinomialnb', MultinomialNB())])\n",
    "\n",
    "\n",
    "    parametros = {\n",
    "        'tfidfvectorizer__stop_words': ['english', stopwords_es, None],\n",
    "        'tfidfvectorizer__max_df': frecuencia_maxima,\n",
    "        'tfidfvectorizer__min_df': frecuencia_minima,\n",
    "        'tfidfvectorizer__ngram_range': ngram_range,\n",
    "        'multinomialnb__alpha': alpha,\n",
    "        'multinomialnb__fit_prior': fit_prior,\n",
    "        'multinomialnb__class_prior': class_prior\n",
    "    }\n",
    "\n",
    "    modelo_rcv = RandomizedSearchCV(modelo, parametros, cv=folds, scoring=metrica)\n",
    "    modelo_rcv.fit(x_train.review_es, y_train_transformed)\n",
    "\n",
    "    dump(modelo_rcv, 'modelos/TP2/nb_hiper.joblib')\n",
    "else: \n",
    "    modelo_rcv = load('modelos/TP2/nb_hiper.joblib')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "edb72575",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-1 {color: black;background-color: white;}#sk-container-id-1 pre{padding: 0;}#sk-container-id-1 div.sk-toggleable {background-color: white;}#sk-container-id-1 label.sk-toggleable__label {cursor: pointer;display: block;width: 100%;margin-bottom: 0;padding: 0.3em;box-sizing: border-box;text-align: center;}#sk-container-id-1 label.sk-toggleable__label-arrow:before {content: \"▸\";float: left;margin-right: 0.25em;color: #696969;}#sk-container-id-1 label.sk-toggleable__label-arrow:hover:before {color: black;}#sk-container-id-1 div.sk-estimator:hover label.sk-toggleable__label-arrow:before {color: black;}#sk-container-id-1 div.sk-toggleable__content {max-height: 0;max-width: 0;overflow: hidden;text-align: left;background-color: #f0f8ff;}#sk-container-id-1 div.sk-toggleable__content pre {margin: 0.2em;color: black;border-radius: 0.25em;background-color: #f0f8ff;}#sk-container-id-1 input.sk-toggleable__control:checked~div.sk-toggleable__content {max-height: 200px;max-width: 100%;overflow: auto;}#sk-container-id-1 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {content: \"▾\";}#sk-container-id-1 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-1 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-1 input.sk-hidden--visually {border: 0;clip: rect(1px 1px 1px 1px);clip: rect(1px, 1px, 1px, 1px);height: 1px;margin: -1px;overflow: hidden;padding: 0;position: absolute;width: 1px;}#sk-container-id-1 div.sk-estimator {font-family: monospace;background-color: #f0f8ff;border: 1px dotted black;border-radius: 0.25em;box-sizing: border-box;margin-bottom: 0.5em;}#sk-container-id-1 div.sk-estimator:hover {background-color: #d4ebff;}#sk-container-id-1 div.sk-parallel-item::after {content: \"\";width: 100%;border-bottom: 1px solid gray;flex-grow: 1;}#sk-container-id-1 div.sk-label:hover label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-1 div.sk-serial::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: 0;}#sk-container-id-1 div.sk-serial {display: flex;flex-direction: column;align-items: center;background-color: white;padding-right: 0.2em;padding-left: 0.2em;position: relative;}#sk-container-id-1 div.sk-item {position: relative;z-index: 1;}#sk-container-id-1 div.sk-parallel {display: flex;align-items: stretch;justify-content: center;background-color: white;position: relative;}#sk-container-id-1 div.sk-item::before, #sk-container-id-1 div.sk-parallel-item::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: -1;}#sk-container-id-1 div.sk-parallel-item {display: flex;flex-direction: column;z-index: 1;position: relative;background-color: white;}#sk-container-id-1 div.sk-parallel-item:first-child::after {align-self: flex-end;width: 50%;}#sk-container-id-1 div.sk-parallel-item:last-child::after {align-self: flex-start;width: 50%;}#sk-container-id-1 div.sk-parallel-item:only-child::after {width: 0;}#sk-container-id-1 div.sk-dashed-wrapped {border: 1px dashed gray;margin: 0 0.4em 0.5em 0.4em;box-sizing: border-box;padding-bottom: 0.4em;background-color: white;}#sk-container-id-1 div.sk-label label {font-family: monospace;font-weight: bold;display: inline-block;line-height: 1.2em;}#sk-container-id-1 div.sk-label-container {text-align: center;}#sk-container-id-1 div.sk-container {/* jupyter's `normalize.less` sets `[hidden] { display: none; }` but bootstrap.min.css set `[hidden] { display: none !important; }` so we also need the `!important` here to be able to override the default hidden behavior on the sphinx rendered scikit-learn.org. See: https://github.com/scikit-learn/scikit-learn/issues/21755 */display: inline-block !important;position: relative;}#sk-container-id-1 div.sk-text-repr-fallback {display: none;}</style><div id=\"sk-container-id-1\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>RandomizedSearchCV(cv=5,\n",
       "                   estimator=Pipeline(steps=[(&#x27;tfidfvectorizer&#x27;,\n",
       "                                              TfidfVectorizer()),\n",
       "                                             (&#x27;multinomialnb&#x27;,\n",
       "                                              MultinomialNB())]),\n",
       "                   param_distributions={&#x27;multinomialnb__alpha&#x27;: [0.1, 0.5, 1.0,\n",
       "                                                                 2.0, 0.8, 0.3,\n",
       "                                                                 1.5],\n",
       "                                        &#x27;multinomialnb__class_prior&#x27;: [None,\n",
       "                                                                       [0.2,\n",
       "                                                                        0.8],\n",
       "                                                                       [0.5,\n",
       "                                                                        0.5]],\n",
       "                                        &#x27;multinomialnb__fit_prior&#x27;: [True,\n",
       "                                                                     False],\n",
       "                                        &#x27;tfidfvectorizer__max_df&#x27;: [0.9, 0.95,\n",
       "                                                                    0.85,\n",
       "                                                                    0.98],\n",
       "                                        &#x27;tfidfvectorizer__min_df&#x27;: [0.05, 0.07,\n",
       "                                                                    0.1, 0.03],\n",
       "                                        &#x27;tfidfvectorizer__ngram_range&#x27;: [(1, 1),\n",
       "                                                                         (1, 2),\n",
       "                                                                         (3,\n",
       "                                                                          1)],\n",
       "                                        &#x27;tfidfvectorizer__stop_words&#x27;: [&#x27;english&#x27;,\n",
       "                                                                        [&#x27;de&#x27;,\n",
       "                                                                         &#x27;la&#x27;,\n",
       "                                                                         &#x27;que&#x27;,\n",
       "                                                                         &#x27;el&#x27;,\n",
       "                                                                         &#x27;en&#x27;,\n",
       "                                                                         &#x27;y&#x27;,\n",
       "                                                                         &#x27;a&#x27;,\n",
       "                                                                         &#x27;los&#x27;,\n",
       "                                                                         &#x27;del&#x27;,\n",
       "                                                                         &#x27;se&#x27;,\n",
       "                                                                         &#x27;las&#x27;,\n",
       "                                                                         &#x27;por&#x27;,\n",
       "                                                                         &#x27;un&#x27;,\n",
       "                                                                         &#x27;para&#x27;,\n",
       "                                                                         &#x27;con&#x27;,\n",
       "                                                                         &#x27;no&#x27;,\n",
       "                                                                         &#x27;una&#x27;,\n",
       "                                                                         &#x27;su&#x27;,\n",
       "                                                                         &#x27;al&#x27;,\n",
       "                                                                         &#x27;lo&#x27;,\n",
       "                                                                         &#x27;como&#x27;,\n",
       "                                                                         &#x27;más&#x27;,\n",
       "                                                                         &#x27;pero&#x27;,\n",
       "                                                                         &#x27;sus&#x27;,\n",
       "                                                                         &#x27;le&#x27;,\n",
       "                                                                         &#x27;ya&#x27;,\n",
       "                                                                         &#x27;o&#x27;,\n",
       "                                                                         &#x27;este&#x27;,\n",
       "                                                                         &#x27;sí&#x27;,\n",
       "                                                                         &#x27;porque&#x27;, ...],\n",
       "                                                                        None]},\n",
       "                   scoring=make_scorer(f1_score))</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-1\" type=\"checkbox\" ><label for=\"sk-estimator-id-1\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">RandomizedSearchCV</label><div class=\"sk-toggleable__content\"><pre>RandomizedSearchCV(cv=5,\n",
       "                   estimator=Pipeline(steps=[(&#x27;tfidfvectorizer&#x27;,\n",
       "                                              TfidfVectorizer()),\n",
       "                                             (&#x27;multinomialnb&#x27;,\n",
       "                                              MultinomialNB())]),\n",
       "                   param_distributions={&#x27;multinomialnb__alpha&#x27;: [0.1, 0.5, 1.0,\n",
       "                                                                 2.0, 0.8, 0.3,\n",
       "                                                                 1.5],\n",
       "                                        &#x27;multinomialnb__class_prior&#x27;: [None,\n",
       "                                                                       [0.2,\n",
       "                                                                        0.8],\n",
       "                                                                       [0.5,\n",
       "                                                                        0.5]],\n",
       "                                        &#x27;multinomialnb__fit_prior&#x27;: [True,\n",
       "                                                                     False],\n",
       "                                        &#x27;tfidfvectorizer__max_df&#x27;: [0.9, 0.95,\n",
       "                                                                    0.85,\n",
       "                                                                    0.98],\n",
       "                                        &#x27;tfidfvectorizer__min_df&#x27;: [0.05, 0.07,\n",
       "                                                                    0.1, 0.03],\n",
       "                                        &#x27;tfidfvectorizer__ngram_range&#x27;: [(1, 1),\n",
       "                                                                         (1, 2),\n",
       "                                                                         (3,\n",
       "                                                                          1)],\n",
       "                                        &#x27;tfidfvectorizer__stop_words&#x27;: [&#x27;english&#x27;,\n",
       "                                                                        [&#x27;de&#x27;,\n",
       "                                                                         &#x27;la&#x27;,\n",
       "                                                                         &#x27;que&#x27;,\n",
       "                                                                         &#x27;el&#x27;,\n",
       "                                                                         &#x27;en&#x27;,\n",
       "                                                                         &#x27;y&#x27;,\n",
       "                                                                         &#x27;a&#x27;,\n",
       "                                                                         &#x27;los&#x27;,\n",
       "                                                                         &#x27;del&#x27;,\n",
       "                                                                         &#x27;se&#x27;,\n",
       "                                                                         &#x27;las&#x27;,\n",
       "                                                                         &#x27;por&#x27;,\n",
       "                                                                         &#x27;un&#x27;,\n",
       "                                                                         &#x27;para&#x27;,\n",
       "                                                                         &#x27;con&#x27;,\n",
       "                                                                         &#x27;no&#x27;,\n",
       "                                                                         &#x27;una&#x27;,\n",
       "                                                                         &#x27;su&#x27;,\n",
       "                                                                         &#x27;al&#x27;,\n",
       "                                                                         &#x27;lo&#x27;,\n",
       "                                                                         &#x27;como&#x27;,\n",
       "                                                                         &#x27;más&#x27;,\n",
       "                                                                         &#x27;pero&#x27;,\n",
       "                                                                         &#x27;sus&#x27;,\n",
       "                                                                         &#x27;le&#x27;,\n",
       "                                                                         &#x27;ya&#x27;,\n",
       "                                                                         &#x27;o&#x27;,\n",
       "                                                                         &#x27;este&#x27;,\n",
       "                                                                         &#x27;sí&#x27;,\n",
       "                                                                         &#x27;porque&#x27;, ...],\n",
       "                                                                        None]},\n",
       "                   scoring=make_scorer(f1_score))</pre></div></div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-2\" type=\"checkbox\" ><label for=\"sk-estimator-id-2\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">estimator: Pipeline</label><div class=\"sk-toggleable__content\"><pre>Pipeline(steps=[(&#x27;tfidfvectorizer&#x27;, TfidfVectorizer()),\n",
       "                (&#x27;multinomialnb&#x27;, MultinomialNB())])</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-3\" type=\"checkbox\" ><label for=\"sk-estimator-id-3\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">TfidfVectorizer</label><div class=\"sk-toggleable__content\"><pre>TfidfVectorizer()</pre></div></div></div><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-4\" type=\"checkbox\" ><label for=\"sk-estimator-id-4\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">MultinomialNB</label><div class=\"sk-toggleable__content\"><pre>MultinomialNB()</pre></div></div></div></div></div></div></div></div></div></div></div></div>"
      ],
      "text/plain": [
       "RandomizedSearchCV(cv=5,\n",
       "                   estimator=Pipeline(steps=[('tfidfvectorizer',\n",
       "                                              TfidfVectorizer()),\n",
       "                                             ('multinomialnb',\n",
       "                                              MultinomialNB())]),\n",
       "                   param_distributions={'multinomialnb__alpha': [0.1, 0.5, 1.0,\n",
       "                                                                 2.0, 0.8, 0.3,\n",
       "                                                                 1.5],\n",
       "                                        'multinomialnb__class_prior': [None,\n",
       "                                                                       [0.2,\n",
       "                                                                        0.8],\n",
       "                                                                       [0.5,\n",
       "                                                                        0.5]],\n",
       "                                        'multinomialnb__fit_prior': [True,\n",
       "                                                                     False],\n",
       "                                        'tfidfvectorizer__max_df': [0.9, 0.95,\n",
       "                                                                    0.85,\n",
       "                                                                    0.98],\n",
       "                                        'tfidfvectorizer__min_df': [0.05, 0.07,\n",
       "                                                                    0.1, 0.03],\n",
       "                                        'tfidfvectorizer__ngram_range': [(1, 1),\n",
       "                                                                         (1, 2),\n",
       "                                                                         (3,\n",
       "                                                                          1)],\n",
       "                                        'tfidfvectorizer__stop_words': ['english',\n",
       "                                                                        ['de',\n",
       "                                                                         'la',\n",
       "                                                                         'que',\n",
       "                                                                         'el',\n",
       "                                                                         'en',\n",
       "                                                                         'y',\n",
       "                                                                         'a',\n",
       "                                                                         'los',\n",
       "                                                                         'del',\n",
       "                                                                         'se',\n",
       "                                                                         'las',\n",
       "                                                                         'por',\n",
       "                                                                         'un',\n",
       "                                                                         'para',\n",
       "                                                                         'con',\n",
       "                                                                         'no',\n",
       "                                                                         'una',\n",
       "                                                                         'su',\n",
       "                                                                         'al',\n",
       "                                                                         'lo',\n",
       "                                                                         'como',\n",
       "                                                                         'más',\n",
       "                                                                         'pero',\n",
       "                                                                         'sus',\n",
       "                                                                         'le',\n",
       "                                                                         'ya',\n",
       "                                                                         'o',\n",
       "                                                                         'este',\n",
       "                                                                         'sí',\n",
       "                                                                         'porque', ...],\n",
       "                                                                        None]},\n",
       "                   scoring=make_scorer(f1_score))"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "modelo_rcv"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "90ee0348",
   "metadata": {},
   "source": [
    "Observamos como se comporta el modelo al estudiarlo a detalle usando la tecnica de random search cross validation "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "c311a7c8",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1-Score del modelo: 0.7728904164999303\n",
      "tfidfvectorizer__ngram_range:(1, 2)\n",
      "tfidfvectorizer__min_df:0.05\n",
      "tfidfvectorizer__max_df:0.85\n",
      "multinomialnb__fit_prior:False\n",
      "multinomialnb__class_prior:[0.5, 0.5]\n",
      "multinomialnb__alpha:0.8\n"
     ]
    }
   ],
   "source": [
    "print(\"F1-Score del modelo: \" + str(modelo_rcv.best_score_))\n",
    "for param, value in modelo_rcv.best_params_.items():\n",
    "    if param == \"tfidfvectorizer__stop_words\":\n",
    "        continue\n",
    "        \n",
    "    print(str(param) + \":\" + str(value))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cb3a755c",
   "metadata": {},
   "source": [
    "Se observa que el modelo tiene un F1-Score bastante parecido a la primera parte del analisis, más sin embargo, el añadido de estos hiperparametros puede ayudar en la simplificación del modelo y mejorar la generalización \n",
    "\n",
    "Entrenamos el modelo con los nuevos hiperparametros y realizamos una validacion cruzada para verificar que no haya ningun problema con el modelo seleccionado "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "d8b14ab9",
   "metadata": {
    "lines_to_next_cell": 2
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfTransformer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfVectorizer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator MultinomialNB from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator Pipeline from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "if not exists('modelos/TP2/nb_conhp.joblib'):\n",
    "    mejores_parametros = modelo_rcv.best_params_\n",
    "\n",
    "    modelo_final = Pipeline(steps=[\n",
    "        ('tfidfvectorizer', TfidfVectorizer(stop_words=mejores_parametros['tfidfvectorizer__stop_words'],\n",
    "                                            max_df=mejores_parametros['tfidfvectorizer__max_df'],\n",
    "                                            min_df=mejores_parametros['tfidfvectorizer__min_df'],\n",
    "                                             ngram_range=mejores_parametros['tfidfvectorizer__ngram_range'])),\n",
    "        ('multinomialnb', MultinomialNB(alpha = mejores_parametros['multinomialnb__alpha'],\n",
    "                                        fit_prior = mejores_parametros['multinomialnb__fit_prior'],\n",
    "                                        class_prior= mejores_parametros['multinomialnb__class_prior']))\n",
    "    ])\n",
    "\n",
    "    modelo_final.fit(x_train.review_es, y_train_transformed)\n",
    "    dump(modelo_final, 'modelos/TP2/nb_conhp.joblib')\n",
    "else:\n",
    "    modelo_final = load('modelos/TP2/nb_conhp.joblib')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "0007ef27",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('modelos/TP2/metricas_nb.joblib'):\n",
    "    kfoldcv =StratifiedKFold(n_splits=10)\n",
    "    resultados_nb = cross_validate(modelo_final, x_train.review_es, y_train, cv=kfoldcv,scoring=metrica ,return_estimator=True)\n",
    "    metricas_nb = resultados_nb['test_score']\n",
    "    modelo_final = resultados_bb['estimator'][np.where(metricas_nb==max(metricas_nb))[0][0]]\n",
    "else: \n",
    "    metricas_nb = load('modelos/TP2/metricas_nb.joblib')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "c4aff267",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Modelo verificado con 10 folds')"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "metric_labelsCV = ['F1 Score']*len(metricas_nb)\n",
    "sns.set_context('talk')\n",
    "sns.set_style(\"darkgrid\")\n",
    "plt.figure(figsize=(6,5))\n",
    "sns.boxplot(metricas_nb)\n",
    "plt.title(\"Modelo verificado con 10 folds\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c6d809a0",
   "metadata": {},
   "source": [
    "Observamos que la validación cruzada reatifica el modelo generado y los hiperparametros usados por lo tanto realizamos la prediccion de peso para la competencia de kaggle \n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "af508411",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('submissions/TP2/NMmejorado8.csv'):\n",
    "    df_submission = pd.DataFrame({'id': review_pruebas['ID'], 'sentimiento': y_pred})\n",
    "    df_submission['sentimiento'] = df_submission['sentimiento'].map({0: 'positivo', 1: 'negativo'})\n",
    "    df_submission.to_csv('NMmejorado8.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0725a219",
   "metadata": {},
   "source": [
    "# Random Forest"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cc5bfce1",
   "metadata": {},
   "source": [
    "Creamnos el modelo con parametros arbitrarios parta obtener una prediccion inicial. Decidimos fijar los valores de samples_leaf y samples_split y n_estimators por nuestra cuenta ya luuego de varioas pruebas obsevamos que con valores por default (1, 2 y 100 respectictivamente) genera un modelo muy grande que no podremos subir a gitthub y cuya efectividad no mejorara mucho respecto a uno un poco mas chico"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "3257af58",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfTransformer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfVectorizer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator DecisionTreeClassifier from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator RandomForestClassifier from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator Pipeline from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "if exists('modelos/TP2/modeloRandomForest-sin-optimizar-final.joblib') == False:\n",
    "    #Creamos un clasificador con hiperparámetros arbitrarios\n",
    "    \n",
    "    rfc = RandomForestClassifier(n_jobs=JOBS,\n",
    "                                 criterion=\"gini\", \n",
    "                                 random_state=SEED, \n",
    "                                 min_samples_leaf=15,\n",
    "                                 min_samples_split=40,\n",
    "                                 n_estimators=40, \n",
    "                                 class_weight=\"balanced\")\n",
    "    \n",
    "    #rfc = RandomForestClassifier(random_state=SEED, n_jobs=-1)\n",
    "    \n",
    "    modeloRandomForest = make_pipeline(TfidfVectorizer(), rfc)\n",
    "\n",
    "    modeloRandomForest.fit(x_train.review_es, y_train)\n",
    "\n",
    "    dump(modeloRandomForest, 'modelos/TP2/modeloRandomForest-sin-optimizar-final.joblib')\n",
    "\n",
    "else:\n",
    "    modeloRandomForest = load('modelos/TP2/modeloRandomForest-sin-optimizar-final.joblib')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f3de93c7",
   "metadata": {},
   "source": [
    "Miramos como fue en test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "be913898",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "    negativo       0.83      0.83      0.83      7300\n",
      "    positivo       0.82      0.82      0.82      7186\n",
      "\n",
      "    accuracy                           0.82     14486\n",
      "   macro avg       0.82      0.82      0.82     14486\n",
      "weighted avg       0.82      0.82      0.82     14486\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(37.08333333333333, 0.5, 'Verdadero')"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "prediccion_rf = modeloRandomForest.predict(x_test.review_es)\n",
    "\n",
    "#performance\n",
    "print(classification_report(y_test,prediccion_rf))\n",
    "\n",
    "\n",
    "#Creamos la matriz de confusión\n",
    "tabla=confusion_matrix(y_test, prediccion_rf)\n",
    "\n",
    "#Grafico la matriz de confusión\n",
    "sns.heatmap(tabla,cmap='GnBu',annot=True,fmt='g')\n",
    "plt.xlabel('Predicho')\n",
    "plt.ylabel('Verdadero')  "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dcb013c7",
   "metadata": {},
   "source": [
    "Hacemos la submission"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "5e8940b1",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('submissions/TP2/randomForest-sin-opti-final.csv'):\n",
    "    prediccionTesteo_rf = modeloRandomForest.predict(review_pruebas.review_es)\n",
    "    df_submission = pd.DataFrame({'id': review_pruebasOriginal['ID'], 'sentimiento': prediccionTesteo_rf})\n",
    "    df_submission.to_csv('submissions/TP2/randomForest-sin-opti-final.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "df23ad27",
   "metadata": {},
   "source": [
    "#### Optimizacion de hiperparametros con Cros Validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "7d6db08a",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfVectorizer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator DecisionTreeClassifier from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator RandomForestClassifier from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator Pipeline from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfTransformer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator GridSearchCV from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'gsrf__criterion': 'gini',\n",
       " 'gsrf__min_samples_leaf': 15,\n",
       " 'gsrf__min_samples_split': 40,\n",
       " 'gsrf__n_estimators': 40,\n",
       " 'tfidfVectorizer__analyzer': 'word',\n",
       " 'tfidfVectorizer__lowercase': True,\n",
       " 'tfidfVectorizer__max_df': 0.1,\n",
       " 'tfidfVectorizer__min_df': 0.01,\n",
       " 'tfidfVectorizer__stop_words': None,\n",
       " 'tfidfVectorizer__strip_accents': None}"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "if exists('modelos/TP2/GsRandomForest_y_tdif-1.joblib') == False:\n",
    "\n",
    "    gsrf = RandomForestClassifier(class_weight=\"balanced\")\n",
    "\n",
    "\n",
    "    modeloRandomForest_cv = Pipeline(steps= [ ('tfidfVectorizer', TfidfVectorizer() ), ('gsrf', gsrf) ] )\n",
    "\n",
    "    param_grid = { \"gsrf__criterion\" : [\"gini\", \"entropy\"], \n",
    "                   \"gsrf__min_samples_leaf\" : [15, 30, 60], #Vamos a hacer muchas combinaciones ya que solo vamos\n",
    "                   \"gsrf__min_samples_split\" : [40, 50, 60],#a correr este modelo 1 sola vez; ya que lo vamos a \n",
    "                   \"gsrf__n_estimators\": [40, 50, 80],  #guardar   \n",
    "                   \n",
    "                    #Veo si conviene usar stopwords o no.\n",
    "                   \"tfidfVectorizer__stop_words\": [None, 'english', stopwords_es], \n",
    "\n",
    "                    \"tfidfVectorizer__strip_accents\": [None, \"ascii\"],\n",
    "                    \n",
    "                    \"tfidfVectorizer__min_df\": [0.01, 0.08, 0.1],\n",
    "\n",
    "                    \"tfidfVectorizer__max_df\" :[ 0.10 , 0.20 , 0.5],\n",
    "\n",
    "                    \"tfidfVectorizer__lowercase\" : [True, False],\n",
    "                    \n",
    "                   \"tfidfVectorizer__analyzer\": [\"word\", \"char\", \"char_wb\"],\n",
    "\n",
    "                    }\n",
    "\n",
    "    #Probamos entrenando sólo con 1 métrica: f1_scoree\n",
    "\n",
    "    rf_gs = GridSearchCV(estimator=modeloRandomForest_cv, param_grid=param_grid, scoring=\"f1\", cv=5, n_jobs=JOBS) #Optimizamos f1_score\n",
    "\n",
    "    #rf_gs = RandomizedSearchCV (estimator=modeloRandomForest_cv, param_grid=param_grid, scoring=\"f1\", cv=5, n_jobs=JOBS) #Optimizamos f1_score\n",
    "\n",
    "    gs_fit = rf_gs.fit(x_train.review_es, y_train)\n",
    "    \n",
    "    #guardamos el grid search\n",
    "    dump(gs_fit, 'modelos/TP2/GsRandomForest_y_tdif-1.joblib')\n",
    "\n",
    "else:\n",
    "    gs_fit = load('modelos/TP2/GsRandomForest_y_tdif-1.joblib')\n",
    "\n",
    "gs_fit.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5472bc41",
   "metadata": {},
   "source": [
    "Podemos ver que para los hiperparametros del RF, Grid Search recomienda son los mismos que los habiamos usado arbitrariamente. Esto se debe a que, Grid search, en general, tiende a elegir\n",
    "el numero mas bajo de sample_leafs, samples_split  y n_estimators. Por ello como ya comentamos tomamos como valores mas chicos los usados en el RF sin optimizar para asegurarnos que en la optimizacion de hiperparametros no escoja valores muy pequeños generando un posible overfitting y un modelo muy complejo. <br>\n",
    "En cuanto a los hiperparametros del vectorizador, recomienda no usar stopwords (las de la lista de la librería NLTK que mandamos como parámetro), no quitar acentos, y usar como analyzer: \"word\". La principal diferencia es en cuanto al minimo y maximo numero de palabras a tener en cuenta para armar el vocabulario. Recomienda eliminar las palabras que representan menos del 1% del vocabulario y aquellas que representan más del 10% del vocabulario."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "dcf21983",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['negativo', 'positivo', 'negativo', ..., 'positivo', 'negativo',\n",
       "       'negativo'], dtype=object)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Obtenemos el mejor modelo\n",
    "mejor_modelo_rf = gs_fit.best_estimator_\n",
    "\n",
    "#Predicción\n",
    "\n",
    "prediccion_mejor_modelo_rf = mejor_modelo_rf.predict(x_test.review_es)\n",
    "prediccion_mejor_modelo_rf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "3ee8188c",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists ('modelos/TP2/modelo_RandomForest_y_tdif-final.joblib'):\n",
    "    dump(mejor_modelo_rf, \"modelos/TP2/modelo_RandomForest_y_tdif-final.joblib\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b740be33",
   "metadata": {},
   "source": [
    "Vemos como le va en test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "824703de",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "    negativo       0.81      0.80      0.81      7300\n",
      "    positivo       0.80      0.81      0.81      7186\n",
      "\n",
      "    accuracy                           0.81     14486\n",
      "   macro avg       0.81      0.81      0.81     14486\n",
      "weighted avg       0.81      0.81      0.81     14486\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(37.08333333333333, 0.5, 'Verdadero')"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#performance\n",
    "print(classification_report(y_test,prediccion_mejor_modelo_rf))\n",
    "\n",
    "\n",
    "#Creamos la matriz de confusión\n",
    "tabla=confusion_matrix(y_test, prediccion_mejor_modelo_rf)\n",
    "\n",
    "#Grafico la matriz de confusión\n",
    "sns.heatmap(tabla,cmap='GnBu',annot=True,fmt='g')\n",
    "plt.xlabel('Predicho')\n",
    "plt.ylabel('Verdadero') "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "40addf37",
   "metadata": {},
   "source": [
    "Vemos que f1_score disminuyo. En el informe profundizamos en las posibles causa de esto."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7363157d",
   "metadata": {},
   "source": [
    "### Cross Validation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "415a24a6",
   "metadata": {},
   "source": [
    "Hacemos Cross validation con 5 folds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "dd6f814f",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfTransformer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfVectorizer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator DecisionTreeClassifier from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator RandomForestClassifier from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator Pipeline from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "kfoldcv =StratifiedKFold(n_splits=5) \n",
    "scorer_fn = make_scorer(sklearn.metrics.f1_score, pos_label='positivo' )\n",
    "\n",
    "if not exists ('modelos/TP2/resultados_cv_randomForest-final'):\n",
    "    resultados_rf = cross_validate(mejor_modelo_rf, x_train.review_es, y_train, cv=kfoldcv,scoring=scorer_fn,return_estimator=True, n_jobs = JOBS)\n",
    "    \n",
    "    dump(resultados_rf, \"modelos/TP2/resultados_cv_randomForest-final\")\n",
    "else:\n",
    "    resultados_rf = load(\"modelos/TP2/resultados_cv_randomForest-final\")\n",
    "\n",
    "metricas_cv_rf = resultados_rf['test_score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "516316e9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Modelo entrenado con 5 folds')"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "metric_labels_CV_rf = ['F1 Score']*len(metricas_cv_rf) \n",
    "sns.set_context('talk')\n",
    "sns.set_style(\"darkgrid\")\n",
    "plt.figure(figsize=(8,8))\n",
    "sns.boxplot(metricas_cv_rf)\n",
    "plt.title(\"Modelo entrenado con 5 folds\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c980b01d",
   "metadata": {},
   "source": [
    "Se puede ver que no hay mucha variacion en los valores obtenidos por lo cual podemos concluir que es un modelo bueno para generalizar."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c134eb89",
   "metadata": {},
   "source": [
    "### Submission Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "9bb7ea57",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('submissions/TP2/randomForest-y_tdif-final-.csv'):\n",
    "    prediccionTesteoMejorRf = mejor_modelo_rf.predict(review_pruebas.review_es)\n",
    "    df_submission = pd.DataFrame({'id': review_pruebasOriginal['ID'], 'sentimiento': prediccionTesteoMejorRf})\n",
    "    df_submission.to_csv('submissions/TP2/randomForest-y_tdif-final.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e38aaef9",
   "metadata": {},
   "source": [
    "La submission nos dio una disminucion de 4 ptos en Kaggle. Al igual que con las metricas en el ocnjunto de test, en el infomre profundizamos en la posibles razones de esto."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9b570db8",
   "metadata": {},
   "source": [
    "# XGBoost\n",
    "\n",
    "Realizamos los mismos ajustes hechos para la busqueda de hiperparametros de NB, en este caso, el modelo XGBoost no es capaz de entender las etiquetas positivo/megativo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "2bcadfb5",
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test_transformed = y_test.map(lambda x: 0 if x == 'positivo' else 1)\n",
    "y_train_transformed = y_train.map(lambda x: 0 if x == 'positivo' else 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5c8253d8",
   "metadata": {},
   "source": [
    "Generamos un modelo base XGBoost que permitira observar un comportamiento global del modelo a la hora de analizar sentimientos en el texto de las reseñas "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "e62557b1",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfTransformer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfVectorizer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator Pipeline from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "if not exists('modelos/TP2/xgb_base.joblib'):\n",
    "    xgb_base = make_pipeline(TfidfVectorizer(), XGBClassifier( random_state=0, n_estimators=100))\n",
    "    xgb_base.fit(x_train.review_es, y_train_transformed)\n",
    "else:\n",
    "    xgb_base = load(\"modelos/TP2/xgb_base.joblib\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ffa5b9be",
   "metadata": {},
   "source": [
    "Evaluando nuestro modelo con las metricas que nos interese"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "d963e8a4",
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = xgb_base.predict(x_test.review_es)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "17eb2c8c",
   "metadata": {},
   "source": [
    "Verificamos la performance del modelo y observamos su matriz de confusión "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "485fa62d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.90      0.93      0.91      7186\n",
      "           1       0.93      0.90      0.91      7300\n",
      "\n",
      "    accuracy                           0.91     14486\n",
      "   macro avg       0.91      0.91      0.91     14486\n",
      "weighted avg       0.91      0.91      0.91     14486\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(37.08333333333333, 0.5, 'Verdadero')"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#performance\n",
    "print(classification_report(y_test_transformed,y_pred))\n",
    "\n",
    "#Creamos la matriz de confusión\n",
    "tabla=confusion_matrix(y_test_transformed, y_pred)\n",
    "\n",
    "#Grafico la matriz de confusión\n",
    "sns.heatmap(tabla,cmap='GnBu',annot=True,fmt='g')\n",
    "plt.xlabel('Predicho')\n",
    "plt.ylabel('Verdadero')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "59b4eee3",
   "metadata": {},
   "source": [
    "Observamos que el modelo tiene un buen desempeño a la hora de\n",
    "\n",
    "Generamos una primera predicción para kaggle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "c4ddbad7",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('submissions/TP2/xgboost_base.csv'):\n",
    "    pred = xgb_base.predict(review_pruebas.review_es)\n",
    "    df_submission = pd.DataFrame({'id': review_pruebasOriginal['ID'], 'sentimiento': pred})\n",
    "    df_submission['sentimiento'] = df_submission['sentimiento'].map({0: 'positivo', 1: 'negativo'})\n",
    "    df_submission.to_csv('submissions/TP2/xgboost_base.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4810dee1",
   "metadata": {},
   "source": [
    "### Busqueda de hiperparametros del XGBoost \n",
    "\n",
    "En un intento de mejorar las metricas de prediccion del modelo, como tambien de simplificarlo y generar un modelo más simple con una capacidad de prediccion similar realizamos la busqueda de hiperparamtros del xgboost utilizando parametros conocidos, como también, modificando hiperparametros en el preprocesamiento del texto para mejorar el analisis de sentimientos y ajustar le posible overfit "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "a9fd1f35",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfVectorizer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator Pipeline from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator TfidfTransformer from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n",
      "/usr/lib/python3.11/site-packages/sklearn/base.py:329: UserWarning: Trying to unpickle estimator RandomizedSearchCV from version 1.2.2 when using version 1.1.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:\n",
      "https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "\n",
    "if not exists('modelos/TP2/hipers_xgb.joblib'):\n",
    "    \n",
    "    estimadores = [90, 100, 110, 150]\n",
    "    frecuencia_maxima = [0.9, 0.95, 0.98]\n",
    "    frecuencia_minima = [0.05, 0.07, 0.03]\n",
    "    profundidad_max = [7, 8, 9, 10, 15]\n",
    "    learning_rate = [0.01, 0.05, 0.1, 0.2]\n",
    "    metrica = make_scorer(sklearn.metrics.f1_score)\n",
    "    folds = 5\n",
    "\n",
    "    parametros = {\n",
    "    'tfidfvectorizer__stop_words': ['english', stopwords_es],\n",
    "    'tfidfvectorizer__max_df': frecuencia_maxima,\n",
    "    'tfidfvectorizer__min_df': frecuencia_minima,\n",
    "    'tfidfvectorizer__ngram_range': [(1,1), (1,2)],\n",
    "    'xgbclassifier__n_estimators': estimadores,\n",
    "    'xgbclassifier__max_depth': profundidad_max,\n",
    "    'xgbclassifier__learning_rate': learning_rate,\n",
    "}\n",
    "\n",
    "    modelo = make_pipeline(TfidfVectorizer(), XGBClassifier())\n",
    "\n",
    "    modelo_rcv = RandomizedSearchCV(modelo, parametros, cv=folds, scoring = metrica)\n",
    "    modelo_rcv.fit(x_train.review_es, y_train)\n",
    "    dump(modelo_rcv, 'modelos/TP2/hipers_xgb.joblib')\n",
    "else: \n",
    "    metrica = make_scorer(sklearn.metrics.f1_score)\n",
    "    modelo_rcv = load('modelos/TP2/hipers_xgb.joblib')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "8e5a2e1c",
   "metadata": {},
   "outputs": [],
   "source": [
    "# print(\"F1-Score del modelo: \" + modelo_rcv.best_score_)\n",
    "# print(\"Hiperparametros del modelo con mejor F1-Score:\" + modelo_rcv.best_params_)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b4585b50",
   "metadata": {},
   "source": [
    "Entrenamos un nuevo modelo usando los hiperparametros encontrados en la busqueda anterior "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "1405c943",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('modelos/TP2/xgb_tuneado.joblib'):\n",
    "    mejores_parametros = modelo_rcv.best_params_\n",
    "\n",
    "    modelo_final = Pipeline(steps=[('tfidfvectorizer', TfidfVectorizer(stop_words=mejores_parametros['tfidfvectorizer__stop_words'],\n",
    "                                                                   max_df=mejores_parametros['tfidfvectorizer__max_df'],\n",
    "                                                                   min_df=mejores_parametros['tfidfvectorizer__min_df'],\n",
    "                                                                   ngram_range=mejores_parametros['tfidfvectorizer__ngram_range'])),\n",
    "                                   ('xgbclassifier', XGBClassifier(n_estimators=mejores_parametros['xgbclassifier__n_estimators'],\n",
    "                                                                  max_depth=mejores_parametros['xgbclassifier__max_depth'],\n",
    "                                                                  learning_rate=mejores_parametros['xgbclassifier__learning_rate']))])\n",
    "\n",
    "    modelo_final.fit(x_train.review_es, y_train_transformed)\n",
    "else:\n",
    "    xgb_tuneado = load('modelos/TP2/xgb_tuneado.joblib')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "49cd0725",
   "metadata": {},
   "source": [
    "Realizo validacion cruzada del modelo para observar su comportamiento *nota* TARDA MUCHISIMO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "d2c3db59",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('modelos/TP2/xgb_metricas.joblib'):\n",
    "    kfoldcv =StratifiedKFold(n_splits=10)\n",
    "    resultados_xgb = cross_validate(xgb_tuneado, x_train.review_es, \n",
    "                                    y_train_transformed, cv=kfoldcv,\n",
    "                                    scoring=metrica ,return_estimator=True)\n",
    "    metricas_xgb = resultados_xgb['test_score']\n",
    "    dump(metricas_xgb, 'modelos/TP2/xgb_metricas.joblib')\n",
    "    xgb_tuneado = resultados_xgb['estimator'][np.where(metricas_xgb==max(metricas_xgb))[0][0]]\n",
    "else: \n",
    "    metricas_xgb = load('modelos/TP2/xgb_metricas.joblib')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "fffe7aad",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Modelo entrenado con 10 folds')"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "metric_labelsCV = ['F1 Score']*len(metricas_xgb)\n",
    "sns.set_context('talk')\n",
    "sns.set_style(\"darkgrid\")\n",
    "plt.figure(figsize=(6,5))\n",
    "sns.boxplot(metricas_xgb)\n",
    "plt.title(\"Modelo entrenado con 10 folds\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "03543fd4",
   "metadata": {},
   "source": [
    "Notas sobre XGBoost\n",
    "\n",
    "El XGBoost más efectivo con un score en kaggle de 0.72 fue generando usando unicamente el conjunto de hiperparametros correspondiente al XGboost y tan solo modificando las stopwords en el procesamiento del texto, sin embargo, se realizo multiples pruebas con multiples hiperparametros y esas busquedas son las reflejadas en el analisis \n",
    "\n",
    "# Red Neuronal\n",
    "\n",
    "## Pre tokenizacion\n",
    "Antes de tokenizar las reviews, vamos a hacer una pequena limpieza adicional.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "926dd61f",
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "if not exists('reviews_filtradas.csv'):\n",
    "    frasesFiltradas = []\n",
    "    for index, value in reviewDf_espanol[\"review_es\"].items():\n",
    "        #Ponemos todas las palabras en lowercase\n",
    "        value = value.lower()\n",
    "\n",
    "        #Saco las stopwords\n",
    "        valueFiltrado = [x for x in value.split() if x not in stopwords_es]\n",
    "        #Vuelvo a unir el texto\n",
    "        valueFiltrado = \" \".join(valueFiltrado)\n",
    "\n",
    "        #Saca los diacriticos de letras como vocales, etc (la ñ se mantiene)\n",
    "        #Expresion regular obtenida de: https://es.stackoverflow.com/a/139811\n",
    "        valueFiltrado = re.sub(r\"([^n\\u0300-\\u036f]|n(?!\\u0303(?![\\u0300-\\u036f])))[\\u0300-\\u036f]+\", r\"\\1\", \n",
    "                                normalize( \"NFD\", valueFiltrado), 0, re.I)\n",
    "        valueFiltrado = normalize( 'NFC', valueFiltrado)\n",
    "\n",
    "        #Saco los signos de puntuacion\n",
    "        #Funcion obtenida de: https://stackoverflow.com/a/266162/13683575\n",
    "        valueFiltrado =  valueFiltrado.translate(str.maketrans('', '', string.punctuation))\n",
    "        valueFiltrado =  valueFiltrado.translate(str.maketrans('', '', '¡'))\n",
    "        valueFiltrado =  valueFiltrado.translate(str.maketrans('', '', '¿'))\n",
    "        \n",
    "        #Anadimos la frase a la lista de frases filtradas\n",
    "        frasesFiltradas.append(valueFiltrado)\n",
    "    reviewDfFiltrado = pd.DataFrame(data={'review_es':frasesFiltradas, 'sentimiento':reviewDf_espanol['sentimiento']})\n",
    "    reviewDfFiltrado.to_csv('reviews_filtradas.csv', index=False)\n",
    "\n",
    "else:\n",
    "    reviewDfFiltrado = pd.read_csv(\"./reviews_filtradas.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "86a8ea27",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>review_es</th>\n",
       "      <th>sentimiento</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>criticos mencionado despues ver solo 1 oz epis...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>pequeña pequeña produccionla tecnica filmacion...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>pense manera maravillosa pasar tiempo fin sema...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>basicamente familia niño pequeño jake piensa z...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>amor tiempo petter mattei pelicula visualmente...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48279</th>\n",
       "      <td>pense pelicula hizo buen trabajo derechano tan...</td>\n",
       "      <td>positivo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48280</th>\n",
       "      <td>mala parcela mal dialogo mala actuacion direcc...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48281</th>\n",
       "      <td>catolica enseñada escuelas primarias parroquia...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48282</th>\n",
       "      <td>voy tener desacuerdo comentario anterior lado ...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48283</th>\n",
       "      <td>nadie espera peliculas star trek altas artes f...</td>\n",
       "      <td>negativo</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>48284 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               review_es sentimiento\n",
       "0      criticos mencionado despues ver solo 1 oz epis...    positivo\n",
       "1      pequeña pequeña produccionla tecnica filmacion...    positivo\n",
       "2      pense manera maravillosa pasar tiempo fin sema...    positivo\n",
       "3      basicamente familia niño pequeño jake piensa z...    negativo\n",
       "4      amor tiempo petter mattei pelicula visualmente...    positivo\n",
       "...                                                  ...         ...\n",
       "48279  pense pelicula hizo buen trabajo derechano tan...    positivo\n",
       "48280  mala parcela mal dialogo mala actuacion direcc...    negativo\n",
       "48281  catolica enseñada escuelas primarias parroquia...    negativo\n",
       "48282  voy tener desacuerdo comentario anterior lado ...    negativo\n",
       "48283  nadie espera peliculas star trek altas artes f...    negativo\n",
       "\n",
       "[48284 rows x 2 columns]"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reviewDfFiltrado"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b7d2b2f6",
   "metadata": {},
   "source": [
    "## Creacion de los sets de entrenamiento"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1f3255ad",
   "metadata": {},
   "source": [
    "Transformamos la columna objetivo de \"positivo\" y \"negativo\" a 0 y 1 para poder usarlos en la red neuronal"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "2d8db401",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 1, 1, ..., 0, 0, 0])"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y = reviewDfFiltrado['sentimiento']\n",
    "y = np.array(list(map(lambda x: 1 if x==\"positivo\" else 0, y)))\n",
    "y"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ec954bcd",
   "metadata": {},
   "source": [
    "Creamos set de train y test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "fdfcb7b8",
   "metadata": {
    "lines_to_next_cell": 2
   },
   "outputs": [],
   "source": [
    "x_train, x_test, y_train, y_test = train_test_split(reviewDfFiltrado[\"review_es\"],\n",
    "                                                    y, \n",
    "                                                    test_size=0.3,  #proporcion 70/30\n",
    "                                                    random_state=SEED) #Semilla 9, como el Equipo !!"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "858bdc46",
   "metadata": {},
   "source": [
    "Creamos el vocabulario en base a nuestro set de train. Las palabras van a estar ordenadas por su frecuencia"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "827cf816",
   "metadata": {},
   "outputs": [],
   "source": [
    "vocabulary = Counter()\n",
    "for fila in x_train:\n",
    "    fila = fila.split()\n",
    "    for i in range(0, len(fila)):\n",
    "        fila[i] = bytes(fila[i], 'utf-8')\n",
    "    vocabulary.update(list(fila))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dede2139",
   "metadata": {},
   "source": [
    "Vamos a ver si el vocabulario se genero correctamente"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "e4c1d989",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(b'si', 24964), (b'solo', 19583), (b'peliculas', 19506), (b'ser', 18017)]"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vocabulary.most_common()[1:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f43ab1d2",
   "metadata": {},
   "source": [
    "Vemos que la palabra mas frecuente es \"pelicula\", cosa que hace sentido.\n",
    "\n",
    "Si no hubiesemos hecho el proceso de remocion de stopwords, es probable que la palabra mas frecuente seria una preposicion"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7e45f898",
   "metadata": {},
   "source": [
    "Para reducir el tamaño del vocabulario, vamos a quedarnos solamente con las 10000 palabras mas frecuentes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "39f99760",
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab_size = 10000\n",
    "truncated_vocabulary = [\n",
    "word for word, count in vocabulary.most_common()[:vocab_size]]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "18b6ce20",
   "metadata": {},
   "source": [
    "Vamos a crear una tabla que asocia cada palabra del diccionario con su coordena (la cual representa su frecuencia en el dataset de entrenamiento)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "6e8ffd82",
   "metadata": {},
   "outputs": [],
   "source": [
    "table = {}\n",
    "for i in range(0, len(truncated_vocabulary)):\n",
    "    table[truncated_vocabulary[i]] = i"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "310c4e61",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{b'pelicula': 0,\n",
       " b'si': 1,\n",
       " b'solo': 2,\n",
       " b'peliculas': 3,\n",
       " b'ser': 4,\n",
       " b'historia': 5,\n",
       " b'tan': 6,\n",
       " b'ver': 7,\n",
       " b'realmente': 8,\n",
       " b'vez': 9,\n",
       " b'mejor': 10,\n",
       " b'bien': 11,\n",
       " b'gran': 12,\n",
       " b'puede': 13,\n",
       " b'hace': 14,\n",
       " b'tiempo': 15,\n",
       " b'despues': 16,\n",
       " b'hecho': 17,\n",
       " b'a\\xc3\\xb1os': 18,\n",
       " b'personajes': 19,\n",
       " b'personas': 20,\n",
       " b'vida': 21,\n",
       " b'final': 22,\n",
       " b'incluso': 23,\n",
       " b'dos': 24,\n",
       " b'parte': 25,\n",
       " b'y': 26,\n",
       " b'como': 27,\n",
       " b'bastante': 28,\n",
       " b'visto': 29,\n",
       " b'asi': 30,\n",
       " b'hacer': 31,\n",
       " b'que': 32,\n",
       " b'nunca': 33,\n",
       " b'podria': 34,\n",
       " b'lugar': 35,\n",
       " b'menos': 36,\n",
       " b'simplemente': 37,\n",
       " b'haber': 38,\n",
       " b'embargo': 39,\n",
       " b'actuacion': 40,\n",
       " b'tipo': 41,\n",
       " b'parece': 42,\n",
       " b'escena': 43,\n",
       " b'buena': 44,\n",
       " b'bueno': 45,\n",
       " b'hombre': 46,\n",
       " b'aqui': 47,\n",
       " b'aunque': 48,\n",
       " b'decir': 49,\n",
       " b'escenas': 50,\n",
       " b'manera': 51,\n",
       " b'sido': 52,\n",
       " b'mientras': 53,\n",
       " b'trabajo': 54,\n",
       " b'cada': 55,\n",
       " b'cualquier': 56,\n",
       " b'el': 57,\n",
       " b'cosas': 58,\n",
       " b'demasiado': 59,\n",
       " b'hizo': 60,\n",
       " b'actores': 61,\n",
       " b'creo': 62,\n",
       " b'personaje': 63,\n",
       " b'primera': 64,\n",
       " b'casi': 65,\n",
       " b'luego': 66,\n",
       " b'director': 67,\n",
       " b'mismo': 68,\n",
       " b'ahora': 69,\n",
       " b'puedo': 70,\n",
       " b'no': 71,\n",
       " b'siempre': 72,\n",
       " b'mal': 73,\n",
       " b'veces': 74,\n",
       " b'trama': 75,\n",
       " b'traves': 76,\n",
       " b'realidad': 77,\n",
       " b'toda': 78,\n",
       " b'debe': 79,\n",
       " b'mayoria': 80,\n",
       " b'todas': 81,\n",
       " b'serie': 82,\n",
       " b'tal': 83,\n",
       " b'momento': 84,\n",
       " b'entonces': 85,\n",
       " b'the': 86,\n",
       " b'muchas': 87,\n",
       " b'esto': 88,\n",
       " b'cine': 89,\n",
       " b'comedia': 90,\n",
       " b'es': 91,\n",
       " b'casa': 92,\n",
       " b'minutos': 93,\n",
       " b'mundo': 94,\n",
       " b'accion': 95,\n",
       " b'tener': 96,\n",
       " b'alguien': 97,\n",
       " b'buen': 98,\n",
       " b'aun': 99,\n",
       " b'ni\\xc3\\xb1os': 100,\n",
       " b'forma': 101,\n",
       " b'mejores': 102,\n",
       " b'real': 103,\n",
       " b'mala': 104,\n",
       " b'musica': 105,\n",
       " b'punto': 106,\n",
       " b'vi': 107,\n",
       " b'amor': 108,\n",
       " b'cuenta': 109,\n",
       " b'nuevo': 110,\n",
       " b'cosa': 111,\n",
       " b'peor': 112,\n",
       " b'general': 113,\n",
       " b'papel': 114,\n",
       " b'especialmente': 115,\n",
       " b'todavia': 116,\n",
       " b'ninguna': 117,\n",
       " b'divertido': 118,\n",
       " b'eso': 119,\n",
       " b'pena': 120,\n",
       " b'original': 121,\n",
       " b'excelente': 122,\n",
       " b'espectaculo': 123,\n",
       " b'probablemente': 124,\n",
       " b'guion': 125,\n",
       " b'ni\\xc3\\xb1o': 126,\n",
       " b'pueden': 127,\n",
       " b'10': 128,\n",
       " b'television': 129,\n",
       " b'dia': 130,\n",
       " b'mujer': 131,\n",
       " b'malo': 132,\n",
       " b'alguna': 133,\n",
       " b've': 134,\n",
       " b'pantalla': 135,\n",
       " b'por': 136,\n",
       " b'joven': 137,\n",
       " b'encuentra': 138,\n",
       " b'razon': 139,\n",
       " b'dvd': 140,\n",
       " b'mayor': 141,\n",
       " b'da': 142,\n",
       " b'deberia': 143,\n",
       " b'completamente': 144,\n",
       " b'horrible': 145,\n",
       " b'siquiera': 146,\n",
       " b'persona': 147,\n",
       " b'interesante': 148,\n",
       " b'se': 149,\n",
       " b'principal': 150,\n",
       " b'dialogo': 151,\n",
       " b'la': 152,\n",
       " b'efectos': 153,\n",
       " b'supuesto': 154,\n",
       " b'guerra': 155,\n",
       " b'hacen': 156,\n",
       " b'tres': 157,\n",
       " b'grandes': 158,\n",
       " b'parcela': 159,\n",
       " b'sino': 160,\n",
       " b'familia': 161,\n",
       " b'dificil': 162,\n",
       " b'idea': 163,\n",
       " b'muestra': 164,\n",
       " b'actor': 165,\n",
       " b'version': 166,\n",
       " b'siendo': 167,\n",
       " b'unica': 168,\n",
       " b'dice': 169,\n",
       " b'ciudad': 170,\n",
       " b'dinero': 171,\n",
       " b'2': 172,\n",
       " b'bajo': 173,\n",
       " b'gente': 174,\n",
       " b'padre': 175,\n",
       " b'trata': 176,\n",
       " b'libro': 177,\n",
       " b'va': 178,\n",
       " b'audiencia': 179,\n",
       " b'puedes': 180,\n",
       " b'gusta': 181,\n",
       " b'unico': 182,\n",
       " b'misma': 183,\n",
       " b'pero': 184,\n",
       " b'noche': 185,\n",
       " b'primer': 186,\n",
       " b'muerte': 187,\n",
       " b'algun': 188,\n",
       " b'humor': 189,\n",
       " b'esposa': 190,\n",
       " b'alli': 191,\n",
       " b'john': 192,\n",
       " b'nueva': 193,\n",
       " b'mas': 194,\n",
       " b'verlo': 195,\n",
       " b'nadie': 196,\n",
       " b'horror': 197,\n",
       " b'pense': 198,\n",
       " b'amigos': 199,\n",
       " b'hombres': 200,\n",
       " b'todo': 201,\n",
       " b'pesar': 202,\n",
       " b'hacia': 203,\n",
       " b'linea': 204,\n",
       " b'seguro': 205,\n",
       " b'quien': 206,\n",
       " b'camara': 207,\n",
       " b'encontrar': 208,\n",
       " b'juego': 209,\n",
       " b'debido': 210,\n",
       " b'lado': 211,\n",
       " b'increible': 212,\n",
       " b'comienza': 213,\n",
       " b'ningun': 214,\n",
       " b'largo': 215,\n",
       " b'ella': 216,\n",
       " b'caracter': 217,\n",
       " b'caso': 218,\n",
       " b'vale': 219,\n",
       " b'sigue': 220,\n",
       " b'problemas': 221,\n",
       " b'terror': 222,\n",
       " b'momentos': 223,\n",
       " b'estilo': 224,\n",
       " b'gusto': 225,\n",
       " b'camino': 226,\n",
       " b'produccion': 227,\n",
       " b'mujeres': 228,\n",
       " b'direccion': 229,\n",
       " b'actuaciones': 230,\n",
       " b'mi': 231,\n",
       " b'haciendo': 232,\n",
       " b'presupuesto': 233,\n",
       " b'ejemplo': 234,\n",
       " b'juega': 235,\n",
       " b'madre': 236,\n",
       " b'finalmente': 237,\n",
       " b'hollywood': 238,\n",
       " b'hicieron': 239,\n",
       " b'agradable': 240,\n",
       " b'problema': 241,\n",
       " b'principio': 242,\n",
       " b'ademas': 243,\n",
       " b'peque\\xc3\\xb1a': 244,\n",
       " b'imagenes': 245,\n",
       " b'quiere': 246,\n",
       " b'resto': 247,\n",
       " b'ni\\xc3\\xb1a': 248,\n",
       " b'suficiente': 249,\n",
       " b'grande': 250,\n",
       " b'terrible': 251,\n",
       " b'absolutamente': 252,\n",
       " b'ir': 253,\n",
       " b'ultima': 254,\n",
       " b'divertida': 255,\n",
       " b'nombre': 256,\n",
       " b'totalmente': 257,\n",
       " b'lleva': 258,\n",
       " b'tratando': 259,\n",
       " b'chico': 260,\n",
       " b'buenos': 261,\n",
       " b'tarde': 262,\n",
       " b'episodio': 263,\n",
       " b'parecia': 264,\n",
       " b'ojos': 265,\n",
       " b'lineas': 266,\n",
       " b'partes': 267,\n",
       " b'grupo': 268,\n",
       " b'definitivamente': 269,\n",
       " b'escuela': 270,\n",
       " b'obtener': 271,\n",
       " b'vea': 272,\n",
       " b'sangre': 273,\n",
       " b'policia': 274,\n",
       " b'amigo': 275,\n",
       " b'peque\\xc3\\xb1o': 276,\n",
       " b'alrededor': 277,\n",
       " b'voz': 278,\n",
       " b'medida': 279,\n",
       " b'asesino': 280,\n",
       " b'cabeza': 281,\n",
       " b'mente': 282,\n",
       " b'hijo': 283,\n",
       " b'fanaticos': 284,\n",
       " b'medio': 285,\n",
       " b'estrellas': 286,\n",
       " b'3': 287,\n",
       " b'en': 288,\n",
       " b'pensar': 289,\n",
       " b'carrera': 290,\n",
       " b'programa': 291,\n",
       " b'imagen': 292,\n",
       " b'basura': 293,\n",
       " b'calidad': 294,\n",
       " b'convierte': 295,\n",
       " b'par': 296,\n",
       " b'aquellos': 297,\n",
       " b'fanatico': 298,\n",
       " b'titulo': 299,\n",
       " b'usted': 300,\n",
       " b'tema': 301,\n",
       " b'dias': 302,\n",
       " b'igual': 303,\n",
       " b'hoy': 304,\n",
       " b'video': 305,\n",
       " b'ellos': 306,\n",
       " b'falta': 307,\n",
       " b'verdadero': 308,\n",
       " b'suficientemente': 309,\n",
       " b'de': 310,\n",
       " b'segunda': 311,\n",
       " b'cerca': 312,\n",
       " b'menudo': 313,\n",
       " b'mitad': 314,\n",
       " b'encontre': 315,\n",
       " b'genial': 316,\n",
       " b'drama': 317,\n",
       " b'of': 318,\n",
       " b'cualquiera': 319,\n",
       " b'buenas': 320,\n",
       " b'decada': 321,\n",
       " b'a\\xc3\\xb1o': 322,\n",
       " b'corazon': 323,\n",
       " b'podia': 324,\n",
       " b'banda': 325,\n",
       " b'aburrido': 326,\n",
       " b'tambien': 327,\n",
       " b'absoluto': 328,\n",
       " b'1': 329,\n",
       " b'parecen': 330,\n",
       " b'tanto': 331,\n",
       " b'hora': 332,\n",
       " b'van': 333,\n",
       " b'especiales': 334,\n",
       " b'elenco': 335,\n",
       " b'importante': 336,\n",
       " b'brillante': 337,\n",
       " b'hermosa': 338,\n",
       " b'primero': 339,\n",
       " b'supongo': 340,\n",
       " b'diferente': 341,\n",
       " b'intenta': 342,\n",
       " b'sr': 343,\n",
       " b'michael': 344,\n",
       " b'viendo': 345,\n",
       " b'quiero': 346,\n",
       " b'reparto': 347,\n",
       " b'posible': 348,\n",
       " b'estrella': 349,\n",
       " b'ambos': 350,\n",
       " b'encanto': 351,\n",
       " b'exito': 352,\n",
       " b'cierto': 353,\n",
       " b'lucha': 354,\n",
       " b'arte': 355,\n",
       " b'dios': 356,\n",
       " b'tierra': 357,\n",
       " b'viaje': 358,\n",
       " b'sexo': 359,\n",
       " b'etc': 360,\n",
       " b'ciertamente': 361,\n",
       " b'intento': 362,\n",
       " b'espectador': 363,\n",
       " b'podrian': 364,\n",
       " b'obra': 365,\n",
       " b'pieza': 366,\n",
       " b'genero': 367,\n",
       " b'inteligente': 368,\n",
       " b'aparentemente': 369,\n",
       " b'deben': 370,\n",
       " b'oportunidad': 371,\n",
       " b'fuerte': 372,\n",
       " b'actriz': 373,\n",
       " b'palabras': 374,\n",
       " b'esta': 375,\n",
       " b'oh': 376,\n",
       " b'creer': 377,\n",
       " b'vista': 378,\n",
       " b'cuanto': 379,\n",
       " b'pasado': 380,\n",
       " b'alla': 381,\n",
       " b'demas': 382,\n",
       " b'dentro': 383,\n",
       " b'donde': 384,\n",
       " b'a': 385,\n",
       " b'extra\\xc3\\xb1o': 386,\n",
       " b'toma': 387,\n",
       " b'este': 388,\n",
       " b'siento': 389,\n",
       " b'miedo': 390,\n",
       " b'uso': 391,\n",
       " b'jovenes': 392,\n",
       " b'junto': 393,\n",
       " b'necesita': 394,\n",
       " b'propia': 395,\n",
       " b'horas': 396,\n",
       " b'principales': 397,\n",
       " b'sabe': 398,\n",
       " b'importa': 399,\n",
       " b'edad': 400,\n",
       " b'o': 401,\n",
       " b'relacion': 402,\n",
       " b'disfrute': 403,\n",
       " b'cantidad': 404,\n",
       " b'historias': 405,\n",
       " b'asesinato': 406,\n",
       " b'ciencia': 407,\n",
       " b'favor': 408,\n",
       " b'opinion': 409,\n",
       " b'hija': 410,\n",
       " b'extremadamente': 411,\n",
       " b'acuerdo': 412,\n",
       " b'obviamente': 413,\n",
       " b'serio': 414,\n",
       " b'modos': 415,\n",
       " b'primeros': 416,\n",
       " b'ultimo': 417,\n",
       " b'and': 418,\n",
       " b'excepto': 419,\n",
       " b'diferentes': 420,\n",
       " b'poder': 421,\n",
       " b'violencia': 422,\n",
       " b'mucha': 423,\n",
       " b'lejos': 424,\n",
       " b'mucho': 425,\n",
       " b'ayuda': 426,\n",
       " b'clase': 427,\n",
       " b'movimiento': 428,\n",
       " b'propio': 429,\n",
       " b'todos': 430,\n",
       " b'sonido': 431,\n",
       " b'tomar': 432,\n",
       " b'clasico': 433,\n",
       " b'verdaderamente': 434,\n",
       " b'dijo': 435,\n",
       " b'pronto': 436,\n",
       " b'comentarios': 437,\n",
       " b'verdad': 438,\n",
       " b'deja': 439,\n",
       " b'saber': 440,\n",
       " b'cara': 441,\n",
       " b'estadounidense': 442,\n",
       " b'varias': 443,\n",
       " b'triste': 444,\n",
       " b'increiblemente': 445,\n",
       " b'atencion': 446,\n",
       " b'detras': 447,\n",
       " b'pobre': 448,\n",
       " b'exactamente': 449,\n",
       " b'tiempos': 450,\n",
       " b'hablar': 451,\n",
       " b'negro': 452,\n",
       " b'ficcion': 453,\n",
       " b'james': 454,\n",
       " b'queria': 455,\n",
       " b'disparo': 456,\n",
       " b'robert': 457,\n",
       " b'supone': 458,\n",
       " b'equipo': 459,\n",
       " b'malos': 460,\n",
       " b'experiencia': 461,\n",
       " b'pasa': 462,\n",
       " b'dio': 463,\n",
       " b'llamado': 464,\n",
       " b'dado': 465,\n",
       " b'david': 466,\n",
       " b'escritor': 467,\n",
       " b'justo': 468,\n",
       " b'escritura': 469,\n",
       " b'desempe\\xc3\\xb1o': 470,\n",
       " b'chica': 471,\n",
       " b'pueda': 472,\n",
       " b'segundo': 473,\n",
       " b'cinematografia': 474,\n",
       " b'sensacion': 475,\n",
       " b'b': 476,\n",
       " b'vuelve': 477,\n",
       " b'5': 478,\n",
       " b'hermano': 479,\n",
       " b'rendimiento': 480,\n",
       " b'fondo': 481,\n",
       " b'aspecto': 482,\n",
       " b'roles': 483,\n",
       " b'monton': 484,\n",
       " b'talento': 485,\n",
       " b'completo': 486,\n",
       " b'interes': 487,\n",
       " b'varios': 488,\n",
       " b'pasar': 489,\n",
       " b'4': 490,\n",
       " b'acerca': 491,\n",
       " b'teatro': 492,\n",
       " b'hecha': 493,\n",
       " b'tampoco': 494,\n",
       " b'ritmo': 495,\n",
       " b'episodios': 496,\n",
       " b'hacerlo': 497,\n",
       " b'cancion': 498,\n",
       " b'nada': 499,\n",
       " b'verdadera': 500,\n",
       " b'peores': 501,\n",
       " b'jugar': 502,\n",
       " b'cuatro': 503,\n",
       " b'mierda': 504,\n",
       " b'completa': 505,\n",
       " b'heroe': 506,\n",
       " b'cambio': 507,\n",
       " b'documental': 508,\n",
       " b'alto': 509,\n",
       " b'esperaba': 510,\n",
       " b'esperando': 511,\n",
       " b'desafortunadamente': 512,\n",
       " b'rapido': 513,\n",
       " b'alta': 514,\n",
       " b'un': 515,\n",
       " b'diferencia': 516,\n",
       " b'nivel': 517,\n",
       " b'sentido': 518,\n",
       " b'viene': 519,\n",
       " b'una': 520,\n",
       " b'padres': 521,\n",
       " b'termina': 522,\n",
       " b'jack': 523,\n",
       " b'palabra': 524,\n",
       " b'ultimos': 525,\n",
       " b'superior': 526,\n",
       " b'perdida': 527,\n",
       " b'pone': 528,\n",
       " b'juntos': 529,\n",
       " b'llena': 530,\n",
       " b'mano': 531,\n",
       " b'los': 532,\n",
       " b'chicos': 533,\n",
       " b'recuerdo': 534,\n",
       " b'encuentran': 535,\n",
       " b'mira': 536,\n",
       " b'esperar': 537,\n",
       " b'esfuerzo': 538,\n",
       " b'blanco': 539,\n",
       " b'reales': 540,\n",
       " b'quizas': 541,\n",
       " b'antes': 542,\n",
       " b'escrito': 543,\n",
       " b'chicas': 544,\n",
       " b'simple': 545,\n",
       " b'mostrar': 546,\n",
       " b'quieres': 547,\n",
       " b'claramente': 548,\n",
       " b'decente': 549,\n",
       " b'jugando': 550,\n",
       " b'malas': 551,\n",
       " b'canciones': 552,\n",
       " b'novela': 553,\n",
       " b'cuerpo': 554,\n",
       " b'pocos': 555,\n",
       " b'80': 556,\n",
       " b'pocas': 557,\n",
       " b'puntos': 558,\n",
       " b'apoyo': 559,\n",
       " b'matar': 560,\n",
       " b'paso': 561,\n",
       " b'media': 562,\n",
       " b'show': 563,\n",
       " b'calificacion': 564,\n",
       " b'fin': 565,\n",
       " b'espero': 566,\n",
       " b'edicion': 567,\n",
       " b'vemos': 568,\n",
       " b'apertura': 569,\n",
       " b'secuencia': 570,\n",
       " b'secuela': 571,\n",
       " b'salir': 572,\n",
       " b'ninguno': 573,\n",
       " b'funciona': 574,\n",
       " b'eventos': 575,\n",
       " b'tv': 576,\n",
       " b'suspenso': 577,\n",
       " b'senti': 578,\n",
       " b'pais': 579,\n",
       " b'duda': 580,\n",
       " b'thriller': 581,\n",
       " b'veo': 582,\n",
       " b'excelentes': 583,\n",
       " b'aburrida': 584,\n",
       " b'perfecto': 585,\n",
       " b'entretenimiento': 586,\n",
       " b'molesto': 587,\n",
       " b'sucede': 588,\n",
       " b'extra\\xc3\\xb1a': 589,\n",
       " b'encanta': 590,\n",
       " b'hijos': 591,\n",
       " b'novia': 592,\n",
       " b'impresionante': 593,\n",
       " b'lee': 594,\n",
       " b'facil': 595,\n",
       " b'convincente': 596,\n",
       " b'cinco': 597,\n",
       " b'diez': 598,\n",
       " b'evitar': 599,\n",
       " b'espectadores': 600,\n",
       " b'pareja': 601,\n",
       " b'recomiendo': 602,\n",
       " b'dan': 603,\n",
       " b'particular': 604,\n",
       " b'sabia': 605,\n",
       " b'basicamente': 606,\n",
       " b'unidos': 607,\n",
       " b'ingles': 608,\n",
       " b'mirar': 609,\n",
       " b'habla': 610,\n",
       " b'sonora': 611,\n",
       " b'muerto': 612,\n",
       " b'george': 613,\n",
       " b'principalmente': 614,\n",
       " b'apenas': 615,\n",
       " b'luz': 616,\n",
       " b'pregunta': 617,\n",
       " b'especial': 618,\n",
       " b'volver': 619,\n",
       " b'elementos': 620,\n",
       " b'facilmente': 621,\n",
       " b'gore': 622,\n",
       " b'vidas': 623,\n",
       " b'mantener': 624,\n",
       " b'hermana': 625,\n",
       " b'musical': 626,\n",
       " b'resultado': 627,\n",
       " b'dejar': 628,\n",
       " b'similar': 629,\n",
       " b'york': 630,\n",
       " b'tienda': 631,\n",
       " b'mensaje': 632,\n",
       " b'ven': 633,\n",
       " b'realista': 634,\n",
       " b'dr': 635,\n",
       " b'peque\\xc3\\xb1os': 636,\n",
       " b'comentario': 637,\n",
       " b'interesantes': 638,\n",
       " b'especie': 639,\n",
       " b'situacion': 640,\n",
       " b'claro': 641,\n",
       " b'material': 642,\n",
       " b'particularmente': 643,\n",
       " b'total': 644,\n",
       " b'generalmente': 645,\n",
       " b'adolescentes': 646,\n",
       " b'oscar': 647,\n",
       " b'20': 648,\n",
       " b'buscando': 649,\n",
       " b'perdido': 650,\n",
       " b'hablando': 651,\n",
       " b'gustan': 652,\n",
       " b'seguir': 653,\n",
       " b'gracias': 654,\n",
       " b'richard': 655,\n",
       " b'uno': 656,\n",
       " b'frente': 657,\n",
       " b'usar': 658,\n",
       " b'viejo': 659,\n",
       " b'observar': 660,\n",
       " b'secuencias': 661,\n",
       " b'cineastas': 662,\n",
       " b'entender': 663,\n",
       " b'sentir': 664,\n",
       " b'ofrece': 665,\n",
       " b'mismos': 666,\n",
       " b'maravilloso': 667,\n",
       " b'representacion': 668,\n",
       " b'diciendo': 669,\n",
       " b'respecto': 670,\n",
       " b'actuar': 671,\n",
       " b'adolescente': 672,\n",
       " b'pensando': 673,\n",
       " b'puntuacion': 674,\n",
       " b'premisa': 675,\n",
       " b'usa': 676,\n",
       " b'aparte': 677,\n",
       " b'diversion': 678,\n",
       " b'estupido': 679,\n",
       " b'relaciones': 680,\n",
       " b'logra': 681,\n",
       " b'divertidos': 682,\n",
       " b'desarrollo': 683,\n",
       " b'peter': 684,\n",
       " b'tonto': 685,\n",
       " b'revision': 686,\n",
       " b'sabes': 687,\n",
       " b'tom': 688,\n",
       " b'mantiene': 689,\n",
       " b'merece': 690,\n",
       " b'feliz': 691,\n",
       " b'animacion': 692,\n",
       " b'vision': 693,\n",
       " b'estadounidenses': 694,\n",
       " b'sorprendio': 695,\n",
       " b'lleno': 696,\n",
       " b'siguiente': 697,\n",
       " b'debil': 698,\n",
       " b'significa': 699,\n",
       " b'llamada': 700,\n",
       " b'entretenido': 701,\n",
       " b'directores': 702,\n",
       " b'dar': 703,\n",
       " b'vivo': 704,\n",
       " b'maravillosa': 705,\n",
       " b'dicho': 706,\n",
       " b'fuerza': 707,\n",
       " b'larga': 708,\n",
       " b'chistes': 709,\n",
       " b'presenta': 710,\n",
       " b'to': 711,\n",
       " b'script': 712,\n",
       " b'jugo': 713,\n",
       " b'drogas': 714,\n",
       " b'predecible': 715,\n",
       " b'broma': 716,\n",
       " b'temporada': 717,\n",
       " b'conjunto': 718,\n",
       " b'hilarante': 719,\n",
       " b'leer': 720,\n",
       " b'risa': 721,\n",
       " b'maestra': 722,\n",
       " b'obtiene': 723,\n",
       " b'imdb': 724,\n",
       " b'lo': 725,\n",
       " b'termino': 726,\n",
       " b'perfectamente': 727,\n",
       " b'anterior': 728,\n",
       " b'sociedad': 729,\n",
       " b'aparece': 730,\n",
       " b'numero': 731,\n",
       " b'corto': 732,\n",
       " b'compa\\xc3\\xb1ero': 733,\n",
       " b'divertidas': 734,\n",
       " b'pude': 735,\n",
       " b'recuerda': 736,\n",
       " b'voy': 737,\n",
       " b'ello': 738,\n",
       " b'vuelta': 739,\n",
       " b'naturaleza': 740,\n",
       " b'estudio': 741,\n",
       " b'70': 742,\n",
       " b'valor': 743,\n",
       " b'comparacion': 744,\n",
       " b'gustaria': 745,\n",
       " b'comenzo': 746,\n",
       " b'lista': 747,\n",
       " b'escenario': 748,\n",
       " b'creditos': 749,\n",
       " b'contrario': 750,\n",
       " b'giro': 751,\n",
       " b'manos': 752,\n",
       " b'local': 753,\n",
       " b'perro': 754,\n",
       " b'rapidamente': 755,\n",
       " b'ex': 756,\n",
       " b'comienzo': 757,\n",
       " b'ves': 758,\n",
       " b'duro': 759,\n",
       " b'aproximadamente': 760,\n",
       " b'suena': 761,\n",
       " b'vivir': 762,\n",
       " b'fue': 763,\n",
       " b'joe': 764,\n",
       " b'baile': 765,\n",
       " b'escuchar': 766,\n",
       " b'corta': 767,\n",
       " b'llega': 768,\n",
       " b'monstruo': 769,\n",
       " b'esperanza': 770,\n",
       " b'razones': 771,\n",
       " b'belleza': 772,\n",
       " b'ideas': 773,\n",
       " b'emocion': 774,\n",
       " b'periodo': 775,\n",
       " b'cree': 776,\n",
       " b'escritores': 777,\n",
       " b'copia': 778,\n",
       " b'espacio': 779,\n",
       " b'salvar': 780,\n",
       " b'basa': 781,\n",
       " b'futuro': 782,\n",
       " b'muestran': 783,\n",
       " b'dejo': 784,\n",
       " b'paul': 785,\n",
       " b'lanzamiento': 786,\n",
       " b'mata': 787,\n",
       " b'leido': 788,\n",
       " b'fan': 789,\n",
       " b'disfrutar': 790,\n",
       " b'escribir': 791,\n",
       " b'semana': 792,\n",
       " b'90': 793,\n",
       " b'conocido': 794,\n",
       " b'muertos': 795,\n",
       " b'darle': 796,\n",
       " b'medico': 797,\n",
       " b'tratar': 798,\n",
       " b'gracioso': 799,\n",
       " b'impresion': 800,\n",
       " b'familiar': 801,\n",
       " b'crimen': 802,\n",
       " b'romance': 803,\n",
       " b'infierno': 804,\n",
       " b'disney': 805,\n",
       " b'cuales': 806,\n",
       " b'perder': 807,\n",
       " b'convirtio': 808,\n",
       " b'oscuro': 809,\n",
       " b'trabajar': 810,\n",
       " b'efecto': 811,\n",
       " b'doy': 812,\n",
       " b'di': 813,\n",
       " b'personal': 814,\n",
       " b'isla': 815,\n",
       " b'posiblemente': 816,\n",
       " b'30': 817,\n",
       " b'actuando': 818,\n",
       " b'agua': 819,\n",
       " b'mirando': 820,\n",
       " b'aire': 821,\n",
       " b'espera': 822,\n",
       " b'promedio': 823,\n",
       " b'con': 824,\n",
       " b'vamos': 825,\n",
       " b'sola': 826,\n",
       " b'baja': 827,\n",
       " b'batalla': 828,\n",
       " b'yo': 829,\n",
       " b'vio': 830,\n",
       " b'nuevamente': 831,\n",
       " b'acabo': 832,\n",
       " b'misterio': 833,\n",
       " b'i': 834,\n",
       " b'aspectos': 835,\n",
       " b'tripulacion': 836,\n",
       " b'comenzar': 837,\n",
       " b'creible': 838,\n",
       " b'secundaria': 839,\n",
       " b'llegar': 840,\n",
       " b'risas': 841,\n",
       " b'bebe': 842,\n",
       " b'sucedio': 843,\n",
       " b'espeluznante': 844,\n",
       " b'villano': 845,\n",
       " b'fantasia': 846,\n",
       " b'is': 847,\n",
       " b'hara': 848,\n",
       " b'humano': 849,\n",
       " b'vive': 850,\n",
       " b'principios': 851,\n",
       " b'jefe': 852,\n",
       " b'algo': 853,\n",
       " b'madera': 854,\n",
       " b'gay': 855,\n",
       " b'miembros': 856,\n",
       " b'antiguo': 857,\n",
       " b'sorpresa': 858,\n",
       " b'maestro': 859,\n",
       " b'caracteres': 860,\n",
       " b'habitacion': 861,\n",
       " b'sexual': 862,\n",
       " b'sra': 863,\n",
       " b'lento': 864,\n",
       " b'william': 865,\n",
       " b'cae': 866,\n",
       " b'obvio': 867,\n",
       " b'mencionar': 868,\n",
       " b'star': 869,\n",
       " b'debo': 870,\n",
       " b'intentan': 871,\n",
       " b'temas': 872,\n",
       " b'ahi': 873,\n",
       " b'deseo': 874,\n",
       " b'crear': 875,\n",
       " b'sale': 876,\n",
       " b'emociones': 877,\n",
       " b'america': 878,\n",
       " b'podemos': 879,\n",
       " b'reir': 880,\n",
       " b'nota': 881,\n",
       " b'probable': 882,\n",
       " b'era': 883,\n",
       " b'decide': 884,\n",
       " b'loco': 885,\n",
       " b'recientemente': 886,\n",
       " b'fiesta': 887,\n",
       " b'enfoque': 888,\n",
       " b'cuando': 889,\n",
       " b'llama': 890,\n",
       " b'unas': 891,\n",
       " b'giros': 892,\n",
       " b'encima': 893,\n",
       " b'debajo': 894,\n",
       " b'lugares': 895,\n",
       " b'finales': 896,\n",
       " b'mezcla': 897,\n",
       " b'sorprendentemente': 898,\n",
       " b'ridiculo': 899,\n",
       " b'siglo': 900,\n",
       " b'in': 901,\n",
       " b'suerte': 902,\n",
       " b'recordar': 903,\n",
       " b'deberian': 904,\n",
       " b'salio': 905,\n",
       " b'espiritu': 906,\n",
       " b'permite': 907,\n",
       " b'jugado': 908,\n",
       " b'humanos': 909,\n",
       " b'directamente': 910,\n",
       " b'12': 911,\n",
       " b'rey': 912,\n",
       " b'tantas': 913,\n",
       " b'altamente': 914,\n",
       " b'cual': 915,\n",
       " b'tomo': 916,\n",
       " b'productores': 917,\n",
       " b'tono': 918,\n",
       " b'color': 919,\n",
       " b'sorprendente': 920,\n",
       " b'dicen': 921,\n",
       " b'preguntas': 922,\n",
       " b'tension': 923,\n",
       " b'poner': 924,\n",
       " b'dolor': 925,\n",
       " b'favorito': 926,\n",
       " b'15': 927,\n",
       " b'odio': 928,\n",
       " b'perfecta': 929,\n",
       " b'mirada': 930,\n",
       " b'cerebro': 931,\n",
       " b'emocionante': 932,\n",
       " b'menor': 933,\n",
       " b'spoilers': 934,\n",
       " b'rara': 935,\n",
       " b'tonta': 936,\n",
       " b'falla': 937,\n",
       " b'tipos': 938,\n",
       " b'dirigido': 939,\n",
       " b'campo': 940,\n",
       " b'mundial': 941,\n",
       " b'antigua': 942,\n",
       " b'emocional': 943,\n",
       " b'cabo': 944,\n",
       " b'terminar': 945,\n",
       " b'hermanos': 946,\n",
       " b'adultos': 947,\n",
       " b'ropa': 948,\n",
       " b'esposo': 949,\n",
       " b'pierde': 950,\n",
       " b'cliche': 951,\n",
       " b'ligeramente': 952,\n",
       " b'ido': 953,\n",
       " b'valores': 954,\n",
       " b'demonios': 955,\n",
       " b'concepto': 956,\n",
       " b'comedias': 957,\n",
       " b'asunto': 958,\n",
       " b'compa\\xc3\\xb1ia': 959,\n",
       " b'acaba': 960,\n",
       " b'sexuales': 961,\n",
       " b'situaciones': 962,\n",
       " b'resulta': 963,\n",
       " b'haberse': 964,\n",
       " b'desnudez': 965,\n",
       " b'scott': 966,\n",
       " b'ejercito': 967,\n",
       " b'queda': 968,\n",
       " b'logro': 969,\n",
       " b'fotografia': 970,\n",
       " b'caja': 971,\n",
       " b'error': 972,\n",
       " b'bill': 973,\n",
       " b'hechos': 974,\n",
       " b'credito': 975,\n",
       " b'epoca': 976,\n",
       " b'sue\\xc3\\xb1o': 977,\n",
       " b'anteriores': 978,\n",
       " b'8': 979,\n",
       " b'detalles': 980,\n",
       " b'proyecto': 981,\n",
       " b'armas': 982,\n",
       " b'golpe': 983,\n",
       " b'lamentablemente': 984,\n",
       " b'parecian': 985,\n",
       " b'excepcion': 986,\n",
       " b'soldados': 987,\n",
       " b'profundidad': 988,\n",
       " b'ama': 989,\n",
       " b'tantos': 990,\n",
       " b'desagradable': 991,\n",
       " b'lider': 992,\n",
       " b'tales': 993,\n",
       " b'centro': 994,\n",
       " b'7': 995,\n",
       " b'busca': 996,\n",
       " b'humana': 997,\n",
       " b'conclusion': 998,\n",
       " b'contiene': 999,\n",
       " ...}"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "table"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c53219df",
   "metadata": {},
   "source": [
    "## Tokenizacion de las reviews"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3a89f8d1",
   "metadata": {},
   "source": [
    "Pasamos de la representacion en texto de la reviews a una version tokenizada. Vamos a usar los ids de la tabla para lograrlo"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "24c9294a",
   "metadata": {},
   "source": [
    "Todas tienen que tener la misma longitud, asique decidimos que todas tengan 300 palabras como maximo. \n",
    "\n",
    "Si faltan palabras o sobran, le ponemos el valor 10001 (para indicar que esta fuera de nuestro vocabulario)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "ca36524c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def pasarDePalabraANumero(conjuntosDePalabras, cantMaxPalabras=300):\n",
    "    listaDeNumeros = []\n",
    "    i=0\n",
    "    for fila in conjuntosDePalabras:\n",
    "        fila = fila.split()\n",
    "        fila = fila[:300]\n",
    "        tokenDeFrase = []\n",
    "        \n",
    "        for word in fila:\n",
    "            tokenDeFrase.append(int(table.get(bytes(word, 'utf-8'),10001)))\n",
    "            \n",
    "        while len(tokenDeFrase) < 300:\n",
    "            tokenDeFrase.append(10001)\n",
    "            \n",
    "        listaDeNumeros.append(np.array(tokenDeFrase))\n",
    "        i+=1\n",
    "    return listaDeNumeros"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "09e322a8",
   "metadata": {},
   "outputs": [],
   "source": [
    "numerico = pasarDePalabraANumero(x_train)\n",
    "numericoTest = pasarDePalabraANumero(x_test)\n",
    "numericoSubmit =  pasarDePalabraANumero(review_pruebas['review_es'])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a05d4829",
   "metadata": {},
   "source": [
    "Ejemplo de como se ven:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "27d393f9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 1502, 10001,    22,  6540,   851,     0,  1111,  2178,    94,\n",
       "         273, 10001,   859,    89,   622,  3346,  2786,  2115,  3613,\n",
       "          27,  1438,     0,    66,  3470,    18,    16,   187,    67,\n",
       "        3613,    27,     3,   622, 10001,   404,   273,  2049,  3240,\n",
       "          68, 10001, 10001, 10001,   172, 10001, 10001,   478, 10001,\n",
       "         184,    85,   178,  2123,   635,   622, 10001,   927, 10001,\n",
       "         346,   440,   384,  5007,   314,   927, 10001,   820,     0,\n",
       "         107,   312,   927, 10001, 10001,    28,   250,   404,   273,\n",
       "         312,   927, 10001,    50, 10001,   580,    28,  3796,   634,\n",
       "        7176,  1181,   201,  1334,   927, 10001,   565, 10001,   119,\n",
       "         761,   801,  2598,    38,   669,   312,   254,   314,     0,\n",
       "          16,    63, 10001,  2883,  1682,  6112,     0, 10001, 10001,\n",
       "        3687,    22,  4457, 10001,   793,    93,  1465,   149, 10001,\n",
       "        6283,   644,   386,    12,   111,   588,   689,   147,    71,\n",
       "         737,    49,   147,  5337,   147,  1664,    94,   277,  2762,\n",
       "        4355,   248,    46,     4,  2727,    55,    46,  1081, 10001,\n",
       "          48, 10001,  2115,   435,   242,     0,  1858,    40,   229,\n",
       "          23,   273, 10001, 10001,  7596,   476,   350,  1437,   249,\n",
       "        5484,  3885,   567,  1054,    20,   265, 10001,   451,   147,\n",
       "         184,  1033,    76,     0,  3140,   372,    59,   132,   324,\n",
       "          49,    68,    16,   945,  1529, 10001,   104,     0,   113,\n",
       "          13,  1063,  4428, 10001,    44, 10001,     9,   889,   723,\n",
       "           0,  9956,  1535,    62,     0,  7100,  3015,  4788, 10001,\n",
       "        8004, 10001,   181,   930,   203,  1059,    50,   797,    44,\n",
       "       10001,  6541,  1290,  3973,   265, 10001,     8,  2303,  3155,\n",
       "         629, 10001, 10001, 10001, 10001,  6227,   131,   631, 10001,\n",
       "         391,   421,    16,   119,    62,     0,   213,  2088, 10001,\n",
       "        3505,  2509,    22,     0,   143,   964,  3883,   650,    98,\n",
       "       10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001,\n",
       "       10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001,\n",
       "       10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001,\n",
       "       10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001,\n",
       "       10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001, 10001,\n",
       "       10001, 10001, 10001])"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "numerico[66]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "f8b877a7",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_list = np.stack(numerico)\n",
    "data_list_test = np.stack(numericoTest)\n",
    "data_list_submit = np.stack(numericoSubmit)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d313eb41",
   "metadata": {},
   "source": [
    "## Creacion de la red (sin hiperparametros)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "03afb764",
   "metadata": {},
   "source": [
    "Vamos a usar la layer intermedia \"GRU\", siguiendo la guia del libro de la bibiliografia. Esta tiene comportamiento muy similar a la LSTM (las cuales son optimas para el analisis de sentimiento)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "781375c2",
   "metadata": {},
   "outputs": [],
   "source": [
    "embed_size = 128\n",
    "redNeuronal = keras.models.Sequential([\n",
    "    \n",
    "keras.layers.Embedding(len(table), embed_size,\n",
    "        input_shape=[None]),\n",
    "    \n",
    "    keras.layers.GRU(128, return_sequences=True),\n",
    "    \n",
    "    keras.layers.GRU(128),\n",
    "    \n",
    "    keras.layers.Dense(1, activation=\"sigmoid\")\n",
    "])\n",
    "redNeuronal.compile(loss=\"binary_crossentropy\", optimizer=\"adam\",\n",
    "            metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "32cf1b85",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding (Embedding)       (None, None, 128)         1280000   \n",
      "                                                                 \n",
      " gru (GRU)                   (None, None, 128)         99072     \n",
      "                                                                 \n",
      " gru_1 (GRU)                 (None, 128)               99072     \n",
      "                                                                 \n",
      " dense (Dense)               (None, 1)                 129       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1,478,273\n",
      "Trainable params: 1,478,273\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "redNeuronal.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "bf4da6d6",
   "metadata": {},
   "outputs": [],
   "source": [
    "if exists('modelos/TP2/redNeuronalSentimiento2.joblib') == False:\n",
    "    historia_modelo=redNeuronal.fit(x = data_list, \n",
    "                                    y = y_train, epochs=4)\n",
    "    dump(redNeuronal, 'modelos/TP2/redNeuronalSentimiento2.joblib')\n",
    "else:\n",
    "    redNeuronal = load('modelos/TP2/redNeuronalSentimiento2.joblib')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "ed5fe1f1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0],\n",
       "       [1],\n",
       "       [0],\n",
       "       ...,\n",
       "       [1],\n",
       "       [0],\n",
       "       [0]])"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "if exists('modelos/TP2/predTestRedSinOptimi.joblib') == False:\n",
    "    y_pred = redNeuronal.predict(data_list_test)\n",
    "    y_predCerteza = np.where(y_pred>0.7,1,0)\n",
    "    dump(y_predCerteza, 'modelos/TP2/predTestRedSinOptimi.joblib')\n",
    "else:\n",
    "    y_predCerteza = load('modelos/TP2/predTestRedSinOptimi.joblib')\n",
    "y_predCerteza"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "30aac623",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ds_validacion=pd.DataFrame(y_predCerteza,y_test).reset_index()\n",
    "ds_validacion.columns=['y_pred','y_real']\n",
    "\n",
    "tabla=pd.crosstab(ds_validacion.y_pred, ds_validacion.y_real)\n",
    "grf=sns.heatmap(tabla,annot=True, cmap = 'Blues', fmt='g')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "28443f93",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.8593124395968521\n",
      "Recall: 0.8075424436404119\n",
      "Precision: 0.8985754103437597\n",
      "f1 score: 0.8506303136909996\n"
     ]
    }
   ],
   "source": [
    "#Calculo las métricas en el conjunto de evaluación\n",
    "accuracy=accuracy_score(y_test,y_predCerteza)\n",
    "recall=recall_score(y_test,y_predCerteza)\n",
    "f1=f1_score(y_test,y_predCerteza,)\n",
    "precision=precision_score(y_test,y_predCerteza)\n",
    "\n",
    "print(\"Accuracy: \"+str(accuracy))\n",
    "print(\"Recall: \"+str(recall))\n",
    "print(\"Precision: \"+str(precision))\n",
    "print(\"f1 score: \"+str(f1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "cc26b4c2",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('submissions/TP2/redesNeuronales2.csv'):\n",
    "    yEnEspanol =y_predCerteza\n",
    "    yEnEspanol = np.array(list(map(lambda x: \"positivo\" if x==1 else \"negativo\", y_predCerteza)))\n",
    "    df_submission = pd.DataFrame({'id': review_pruebasOriginal['ID'], 'sentimiento': yEnEspanol})\n",
    "    df_submission.to_csv('submissions/TP2/redesNeuronales2.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "100c0b61",
   "metadata": {},
   "source": [
    "## Busqueda de hiperparametros\n",
    "\n",
    "Vamos a realizar la busqueda de hiperparametros:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "c11c4949",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss='binary_crossentropy'\n",
    "metrics=['accuracy']\n",
    "optimizer=\"adam\"\n",
    "\n",
    "def creador_modelo(learning_rate = 0.1, \n",
    "                   activation = 'sigmoid', \n",
    "                   output = 2, \n",
    "                  hidden_layers = 2\n",
    "                  ):\n",
    "    modeloHiper = keras.Sequential()\n",
    "    \n",
    "    modeloHiper.add(keras.layers.Embedding(len(table), embed_size,\n",
    "        input_shape=[None]))\n",
    "    \n",
    "    modeloHiper.add(keras.layers.GRU(128, return_sequences=True))\n",
    "    modeloHiper.add(keras.layers.GRU(128))\n",
    "    \n",
    "    for i in range(hidden_layers):\n",
    "        modeloHiper.add(keras.layers.Dense(output, activation=activation))\n",
    "\n",
    "    modeloHiper.add(keras.layers.Dense(1, activation=\"sigmoid\"))\n",
    "    \n",
    "    modeloHiper.compile(\n",
    "      optimizer=optimizer,\n",
    "      loss=loss, \n",
    "      metrics=metrics, \n",
    "    )\n",
    "    return modeloHiper\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "f2b2f774",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_12350/2109703127.py:1: DeprecationWarning: KerasClassifier is deprecated, use Sci-Keras (https://github.com/adriangb/scikeras) instead. See https://www.adriangb.com/scikeras/stable/migration.html for help migrating.\n",
      "  model = KerasClassifier(build_fn=creador_modelo,\n"
     ]
    }
   ],
   "source": [
    "model = KerasClassifier(build_fn=creador_modelo, \n",
    "                        verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "2a074f48",
   "metadata": {},
   "outputs": [],
   "source": [
    "param_grid = { \n",
    "                  \"hidden_layers\" : [1, 5, 10, 15, 20], \n",
    "                    \"output\" : [1, 2, 4, 8, 32, 64], \n",
    "                    \"batch_size\" : [5, 10, 20],\n",
    "                    \"epochs\" : [5, 10, 15],\n",
    "                   \"activation\": [\"sigmoid\", \"relu\", \"softmax\", \"softplus\", \"elu\", ]\n",
    "             } "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6e551335",
   "metadata": {},
   "source": [
    "Vamos a realizar 5 random searches para buscar los mejores hiperparametros para la red"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "4ab2172f",
   "metadata": {},
   "outputs": [],
   "source": [
    "rs = RandomizedSearchCV(estimator=model, \n",
    "                        param_distributions=param_grid,\n",
    "                        n_jobs=JOBS, cv=3,\n",
    "                        n_iter=8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "f650fc3e",
   "metadata": {},
   "outputs": [],
   "source": [
    "if exists('modelos/TP2/rs_redNeuronal.joblib') == False:\n",
    "    rs_redNeuronal=rs.fit(X = data_list, \n",
    "                            y = y_train)\n",
    "\n",
    "    dump(rs_redNeuronal, 'modelos/TP2/rs_redNeuronal.joblib')\n",
    "else:\n",
    "    rs_redNeuronal = load('modelos/TP2/rs_redNeuronal.joblib')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "1d70a5ad",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'output': 4,\n",
       " 'hidden_layers': 1,\n",
       " 'epochs': 10,\n",
       " 'batch_size': 5,\n",
       " 'activation': 'softplus'}"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rs_redNeuronal.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "23c3bc55",
   "metadata": {},
   "source": [
    "## Creamos la red con los hiperparametros encontrados arriba"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "cc732743",
   "metadata": {},
   "outputs": [],
   "source": [
    "embed_size = 128\n",
    "modeloHiper = keras.models.Sequential()\n",
    "    \n",
    "modeloHiper.add(keras.layers.Embedding(len(table), embed_size,\n",
    "        input_shape=[None]))\n",
    "    \n",
    "modeloHiper.add(keras.layers.GRU(128, return_sequences=True))\n",
    "    \n",
    "modeloHiper.add(keras.layers.GRU(128))\n",
    "    \n",
    "for i in range(rs_redNeuronal.best_params_['hidden_layers']):\n",
    "    modeloHiper.add(keras.layers.Dense(rs_redNeuronal.best_params_['output'],\n",
    "                                       activation=rs_redNeuronal.best_params_['activation']))\n",
    "    \n",
    "modeloHiper.add(keras.layers.Dense(1, activation=\"sigmoid\"))\n",
    "\n",
    "modeloHiper.compile(loss=\"binary_crossentropy\", optimizer=\"adam\",\n",
    "            metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "88588004",
   "metadata": {},
   "outputs": [],
   "source": [
    "if exists('modelos/TP2/redNeuronalHiper.joblib') == False:\n",
    "    historia_modelo=modeloHiper.fit(x = data_list, \n",
    "                                    y = y_train, epochs=rs_redNeuronal.best_params_['epochs'])\n",
    "\n",
    "    dump(modeloHiper, 'modelos/TP2/redNeuronalHiper.joblib')\n",
    "\n",
    "else:\n",
    "    modeloHiper = load('modelos/TP2/redNeuronalHiper.joblib')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "88746e80",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0],\n",
       "       [1],\n",
       "       [0],\n",
       "       ...,\n",
       "       [0],\n",
       "       [0],\n",
       "       [0]])"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "if exists('modelos/TP2/predTestRedConOptimi.joblib') == False:\n",
    "    y_pred = modeloHiper.predict(data_list_test)\n",
    "    y_pred\n",
    "    y_predCerteza = np.where(y_pred>0.7,1,0)\n",
    "    y_predCerteza\n",
    "    dump(y_predCerteza, 'modelos/TP2/predTestRedConOptimi.joblib')\n",
    "else:\n",
    "    y_predCerteza = load('modelos/TP2/predTestRedConOptimi.joblib')\n",
    "y_predCerteza"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "2008db64",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlIAAAHHCAYAAAB0nLYeAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABiu0lEQVR4nO3deVxUVePH8c+wCCgibuCGC+6aZmmQZK6pZZvikkvlTy21bLFNzTXXrCezp8zUrEw0U1MjyVxyy73MfQUSV1REBVE2Geb3B808IQPCODpMft/Pixd07zn3nMuj8uWcc881mEwmEyIiIiJSYC6O7oCIiIiIs1KQEhEREbGRgpSIiIiIjRSkRERERGykICUiIiJiIwUpERERERspSImIiIjYSEFKRERExEZuju6As6n40jJHd0Gk0Fn2TmtHd0GkUAoKLHHb2/C67xW7XCdl9zS7XOduoxEpERERERtpREpERMSZGTQm4kgKUiIiIs7MYHB0D+5qirEiIiIiNtKIlIiIiDPT1J5DKUiJiIg4M03tOZRirIiIiIiNNCIlIiLizDS151AKUiIiIs5MU3sOpRgrIiIiYiONSImIiDgzTe05lIKUiIiIM9PUnkMpSImIiDgzjUg5lL77IiIiIjbSiJSIiIgz09SeQylIiYiIODNN7TmUvvsiIiIiNtKIlIiIiDPT1J5DKUiJiIg4M03tOZS++yIiIiI20oiUiIiIM9OIlEMpSImIiDgzF62RciTFWBEREREbaURKRETEmWlqz6EUpERERJyZtj9wKAUpERERZ6YRKYfSd19ERETERhqREhERcWaFeGrv0qVLfPnll6xbt47Y2Fg8PT1p2LAhL7zwAk2bNs1RfuvWrXz55ZccOXKE1NRUAgMD6d69O126dMFg5T4zMjJYsmQJCxcu5Pjx47i5udGgQQP69+9PcHCw1T4lJiYye/Zs1qxZQ2xsLD4+PjRr1oxBgwYREBBQ4HvUiJSIiIgzM7jY58POjh07xtNPP83XX3/N9evXadGiBRUrVmTz5s306dOH1atXZyu/YMEC+vTpwx9//EG9evUIDg7mr7/+YuTIkQwfPjzH9U0mE0OHDmX06NGcPn2akJAQatWqxdatW+nduzdLlizJUSchIYGePXsya9YsjEYjLVu2pGTJkixbtoyOHTty5MiRAt+nRqRERETErjIyMnjzzTeJi4vj+eefZ9iwYbi6ugLw448/MnToUIYNG0aLFi3w8PAgJiaG8ePH4+3tTVhYGPXq1QMgNjaW3r17s3TpUlq0aMGjjz5qaWPJkiVERERQt25d5syZg6+vLwDbtm1jwIABjB07lpCQEMqXL2+pM2nSJKKjowkNDWXChAmWPs2YMYOpU6fyzjvvEB4ejotL/oOlRqREREScmcFgnw87WrNmDYcPH6Zx48YMHz7cElgAOnbsSPPmzSlVqhSHDx8GYPbs2RiNRvr162cJUQAVKlRg9OjRljL/NHPmTABGjhxpCVEATZs2pXfv3qSlpREWFmY5fvr0aSIiIvDx8WHEiBHZ+jRw4EAaNWpEZGQkmzZtKtC9KkiJiIg4s0I4tffLL78A0K9fP6trm7788kt+/fVXGjVqBMD69esBaNeuXY6yISEhFC9enP3793P+/HkAoqOjOXnyJKVLl6Zx48Y56rRv3x6AtWvXWo5t3LgRo9FISEgI3t7e+aqTH5raExEREbs6cOAAAPfddx8JCQmsWLGCI0eO4ObmRpMmTWjfvr1lRCg+Pp6LFy/i7u5OYGBgjmu5uroSGBjI3r17OXr0KP7+/kRGRgJQs2ZNq0HNfPzkyZOkpKTg5eVlqVOrVi2rfa5RowZAgddJKUiJiIg4s0L21F56ejpnzpzBzc2No0eP8sYbb3D58mXL+fnz51OvXj1mzJiBv78/cXFxAJQpUybXtUl+fn4AlrLmz/7+/lbLe3h44OPjQ2JiIvHx8QQEBNy0zo1t5JeClIiIiDOz07Tc999/z6JFiwpUp1u3bnTv3j3bsatXrwJZT9W9/PLLNGjQgCFDhlCtWjUiIyOZOHEi+/fv56WXXmLx4sUkJycD4OnpmWs7Hh4eAJay165dy3cdc1lzXS8vL6vlzdcyl8svBSkRERHhwoULHDx4sMB1bpSeng6A0WikUqVKfPXVV7i7uwNZU33ffPMNjz32GAcPHmT16tWWESJrU3Q3MplMAJZpwfzUMctvHXMb+aUgJSIi4szsNLVXtmxZ6tevX+A6N/rnKFGPHj0sIcqsePHiPPnkk3z99dds27aNXr16AZCampprO2lpaQAULVoUgGLFiuW7jnkEylwnJSXFannztXIbscqNgpSIiIgzs9PUXvfu3XNM09miePHiFClShPT0dCpVqmS1jPn45cuXLSNS8fHxmEwmqyNG5nVL5nVM5jrWRsQgK0RduXIFg8FgCXs3q3NjG/ml7Q9EREScWSHb/sDV1ZWaNWsCWLYruFF8fDwApUqVwtfXF39/f9LT0zlx4kSOskajkWPHjgFQu3btbJ+joqKsXj8qKgqTyURAQIBlFOtmdaKjo7OVyy8FKREREbGrli1bAvDTTz/lOGcymfjtt98ALO/Da9GiBUCO18YAbNmyhaSkJOrUqUO5cuUAqFKlClWrViUuLo49e/bkqLNy5cps/QBo3rw5Li4ubN682eqC8lWrVgHQqlWrfN5lFgUpERERZ1YIdzbv3r07Pj4+/P7770ybNs2ygNtkMvHpp59y4MABAgICeOSRRwDo1asXrq6uzJw5M1swio2NZfz48QAMGDAgWxvPP/88AKNGjbKMcEHWK2Lmzp2Lu7s7ffv2tRz39/enbdu2JCQkMGbMGMuieMDSbo0aNWjTpk2B7tVgKujy9LtcxZeWOboLIoXOsndaO7oLIoVSUGCJ296G19Mz7XKdlPABNy9UAL/99huvvvoqqampVK5cmdq1axMZGcmJEycoUaIEs2fPpmHDhpbys2bNYsqUKbi6uhIUFISnpyc7duwgOTmZrl27MmHChGzXz8zM5KWXXmLDhg14e3sTHBxMUlISO3fuxGQyMXnyZDp27JitTlxcHD179uTUqVNUqFCBBg0aEBMTQ2RkJD4+PsybN6/AU3sKUgWkICWSk4KUiHV3c5ACOHHiBDNmzGDr1q1cvHiRMmXK0KxZMwYMGEBAQECO8mvXrmXOnDkcPHgQg8FAtWrV6NGjB506dbK6Wef169eZN28eS5cu5cSJE3h7e1O3bl0GDBhAUFCQ1T5dunSJ6dOns27dOuLi4vDz8yMoKIhBgwZZ7dPNKEgVkIKUSE4KUiLW3ZEg1XGWXa6T8mN/u1znbqPtD0RERJyZnV84LAWj776IiIiIjTQiJSIi4swK2UuL7zYKUiIiIk6sIO+bE/vT1J6IiIiIjTQiJSIi4sQ0IuVYClIiIiLOTDnKoRSkREREnJhGpBxLa6REREREbKQRKRERESemESnHUpASERFxYgpSjqWpPREREREbaURKRETEiWlEyrEUpERERJyZcpRDaWpPRERExEYakRIREXFimtpzLAUpERERJ6Yg5Via2hMRERGxkUakREREnJhGpBxLQUpERMSJKUg5loKUiIiIM1OOciitkRIRERGxkUakREREnJim9hxLQUpERMSJKUg5lqb2RERERGykESkREREnphEpx1KQEhERcWbKUQ6lqT0RERERG2lESkRExIlpas+xFKREREScmIKUY2lqT0RERMRGGpESERFxYhqRciwFKRERESemIOVYClIiIiLOTDnKobRGSkRERMRGGpESERFxYpracywFKRERESemIOVYClJyW2yf0I6A0sXyLFP3zQiupFy3/Hd1f28GPFKTh2qXoZyvFxnGTGLirvHLnlhmr/uLa2kZlrKVShVlx8T2+e7PlIjDfPzzkTzLNKjsy/IhLXB3daHiS8vyfW2Rgrp04Twrls5n/5/biY87C0BZ/wo0CmrGY517UcK3VI46SVcSCF/wNbu2/8al+DiKeftQq15DHu/yHDXqNsi1rb+OHmTVj98TeXAPCZcvUqSIBxUrV+PBlu1p0yEUN3d3q/Wup6ezZvlidvy2hrOnT5CRkYFf+Yo82KItHUJ7UcTD0z7fDBEnpyAldlfc042A0sXIMGayK+ZyruUyMjMtXz9xfwX++39N8HR3Je26kZgL1yjm4Ur9SiVoUNmXzsEBPPPJZs4mpAKQlmHk9+iLefajrI8H1fy8AYiJu5pn2SJuLvy3d2PcXbVsUG6vowd28/HYt0m+moSLiyv+FSqRmZnJ2dMnOXMyjM1rVzBk4qdUrlbTUifx8kXGvfUicWdPU8TDk8rVanApPo6dWzewa/sm+r72Li3aP5WjrVU/fs/8Lz/BlJlJEQ8PKgRU5eqVRKKPHCD6yAG2rV/F0Imf4lXMO1u9xMsX+WDEq5yKicbFxZXyAVVIS03hzIljLJk7kz+3buTdydMpekM9cQyNSDmWgpTYXb1KJQCIibtGpym/3bR85TJFLSFq8faTjFq4l6TUrNGn+pVKMK1vE2qV92HGC0E8/VHW9S5cScvz2p7uLvw8rBUA3205zrI/TufZh6FP1aN2BZ983Z+Ira5dTeK/E4aRfDWJhk2a8uIbo/AtVQaAuLNnmPnRe0Qe2ssn495h8syFFCniAcC090cQd/Y099wXzCvvTqRYcR8yMzNZ8UMYC7/5nG+mTaZG3QZUrFzN0lbkwb3MnzUVk8nE412eo/NzA3AvUgSAQ3t3MuM/Y/jr6AG+/mwyg4ZNsNQzmUx8Nmk4p2KiqVi5Gq+P+pDylaoAcGT/Lj6bNJzj0UdYNGc6/zdoyJ361klelKMcSr9+i92Zg9SR2MR8le/Xqjqe7q4cOJXAm3P/tIQogIOnE3lh5g4yjJk0qV6aB6qXztc1x3W7lzoVfIg+l8SohfvyLNsksBQvtqlB8j+mDkVuh01rIkhKvEzJ0mV55d1JlhAF4Fe+Iq+P+pBi3j5cOBfLH5vXAXB4358c2b8LT6+ivDx0PMWKZwV+FxcXnujWm5BWj2LMyOCn77/J1tbPS8IwmUzcF/ww3fu9aglRAPXubUL/t8YAsH3jai5eOG859/vmtRw9sBuvosUY9v7nlhAFUKfB/TzTZxAAm3/9mYwM/Z0RUZASu/tfkLqSr/IhtcoCsPzPM2Sacp7/6/xVos9nTc01quJ70+s1q12WXs2qkplp4u15u0m9bsy1rFcRVz7p3ZhMk4mPlh/OV39FbHV4358ANApqhlfRnGsIfXxLUrNe1nqnY5GHAPhtTQQA9z/YnOIlfHPUafN4ZwD+3LaR9LTU/7W1N6utpi3bWe1L/UYP4OmV1YeYv9sC2LTmZwA6dH42W9Aze+ChVoQ+25/u/V4l43p6Hncrd4rBYLDLh9hGU3tid/Uq/h2kzuQvSI1etI+AMkXZncd6KvNfcReXvP+yuxjgva5ZP4iW/H6KP/7Kex3VqNB7qObnzScrjrDvZEK++itiq6e79+WBZq0pV7FyrmVMf/8ykfn3GsLow/sBqFX/XqvlA2vXx9XVlbTUFGKiDlP7nvvIzMzklXcncfHCeWrXb5RLO//7rcXcVqbRyKG9OwFo8lBLq/W8innTqdcLufZf7jyFIMdSkBK7cjFA7QrFATifmMqLrasTVKMMPl5unEtIZe2BcyzfdYZ//BvOtqh4tkXlfs1a5Ytb1i9Fnk3Ks/1ezapSt2IJUq8b+SD8YJ5lH65Tlucersah04l8suIITQLzN20oYqvA2vUIrF0v1/NJiQmWUauKlauRmZnJhXNnAPArX8lqHTc3N0qW9iM+7ixnz5yk9j334eLiQsMmTfPsy76d20hNuZbVVpVAAM7FnuJ6ehouLq5UqFSVKwmX2fTrz0Qd2ktqagrlKgTQrE2HPJ8SFLnbKEiJXQX6F8erSNYfqwWvPURxr+yPVnd5sDIvxlyi34ztxF1Ju+n1XAwwqXsjAM5eTmHL0Qt5ln25XS0Afth+0vKEnzXenm5Mee5+MjJNDP72T64brcwpitxhYTOmkJ6WShEPTx5o1pprV69gNGZNTfuUKJlrPW+fEsTHneVqYkK+2klNSea7Lz8BoFrNupZF6uatGIp5F+fg3p1M/2AkV6/8b63jwd2/s/bnJTzaqQc9XxyskZBCQv8/OJbTBKmkpCQ2bdrEkSNHiIuLIzk5GZPJhJeXF35+ftSuXZuHHnqIUqVy7r8id069Sv978m3victM+fkI+04k4OnuQqv6/ozq3ID7q5Ui7JUQnvhgw00DzOSe99G0VtY6jXFL95OekZlr2Q73VaBymWIYM03MWJPHEBcwvltDKpYqypSIwxw8nb9F8SK3048LvmLbhlUAdOrZjxK+pbItAv/nYvEbmZ/uS0u7+S8nGdev89mk4Zw9fQIXF1eeHfCm5VxqcvLf10nhv+OH4Fe+Iq+8O4ma9RqSfDWJ9b/8yLLvZrNy2QJK+JbiiW69bbpXsa/CGKRiYmJ49NFH8yyzbdu2bD+zt27dypdffsmRI0dITU0lMDCQ7t2706VLF6v3mJGRwZIlS1i4cCHHjx/Hzc2NBg0a0L9/f4KDg622mZiYyOzZs1mzZg2xsbH4+PjQrFkzBg0aREBAgE33WuiD1LVr15g6dSqLFi3i+vXr2eb1/8lgMODu7k7Xrl158803KVYs780g5faIvZTCV+v/IjPTxNgl+y1TeKnXjSz74zR7TySwangr7gnwpcdDVZn7W4zV67gY4INe99HzoaoAfLX+L37aeSbPtvu1qgHAyj2xxFy4lmu5tg3K0a1pFQ6cSuDTX44W/CZF7Gzp/C9ZNu9LIGtR+eNdnweynswzy+uHpYmsv2iGm6whTE9L5dOJ77Jv51YAuvd7Jdvaq/T0tL/LpVHGrzwjPpxJMe+sqfoipTzo1OsFMjMz+fG72fy44GtaPdbJ8hShOFDhy1EcOpT1AEONGjWoW7eu1TIeHh6WrxcsWMB7772Hu7s7wcHBuLu7s337dkaOHMmuXbt4//33s9U1mUwMHTqUiIgISpQoQUhICAkJCWzdupUtW7YwceJEOnfunK1OQkICvXr1Ijo6msqVK9OyZUtiYmJYtmwZa9asYf78+dSpU6fA91qog1RKSgq9evXi6NGjeHp6EhISQs2aNfHz88PTM2tX3dTUVOLi4oiKimLHjh1899137Nmzh7lz5ypMOcDOY5fYeexSruePxV1l6e+nePbhajx6b3mrQcqriCuf932A9veWB2Dh1hOMWZz3FgblfT1pEpj1m80PO07lWq5ksSJ80Os+0jMyeePbP8mw9pigyB1iNGYwd/p/WLciayf9Bo0fZNC7Ey2hydPTy1LWHHKsuZ6e9fSceWTKmsSES3wy9m2ijxwAoGPPF3gstFe2Mh7/2K28fcfulhD1T092e56fF88lLTWFg3v+IOjhNje7TbkLHTiQ9efs+eef55lnnsmzbExMDOPHj8fb25uwsDDq1ctaRxgbG0vv3r1ZunQpLVq0yDbCtWTJEiIiIqhbty5z5szB19cXyBrlGjBgAGPHjiUkJITy5ctb6kyaNIno6GhCQ0OZMGECrq6uAMyYMYOpU6fyzjvvEB4enu0XmPwo1EFqxowZHDlyhNatWzNp0iTLNyo3CQkJDB8+nPXr1/PNN9/wyiuv3JmOSoGYp9Iql8kZdP1LePLty01pUNkXgFm/RjN2yf6bXrP9vRVwcTFwJeU6Gw6dz7XcpO734l/Ck//8dIhD+XyqUOR2SLl2lU8nvsuB3TsACG7eloFvv5ftlS0eXkVxdy/C9evp2dYq3SjpSgIAPlZeLQNw5mQMH40eTPz5sxgMBnq8+DqPdeqZo9w/dyqvUr2W1WsV8fDEr3xFzpyMIe5c3qPEcmcUxqk984hUgwY3fzBh9uzZGI1G+vXrZwlRABUqVGD06NG88MILzJ49O1uQmjlzJgAjR47Mlg2aNm1K7969mTVrFmFhYQwZkrVp7OnTp4mIiMDHx4cRI0ZYQhTAwIEDWb9+PXv27GHTpk20aNGiQPdaqPeRWrlyJX5+fnzyySc3DVEAvr6+TJ06lbJly7Jy5crb30GxymAAd9fc/2Kb/85fN2Zf71S7QnEihrSgQWVfjJkmRi3am68QBVhGr1btPZvnOqqnmmQ9+fTOU/U480WnbB8/vPmwpZz5WLcHc39MXcRWly6cZ9zbL1pCVIcuzzJo2IQc771zcXGhXKWsP4MXzsdavVZGRgYJF7MewrC2rcLhfX8y7q0XiD9/FvciHrzy7iSrIQqgwj92Rr+exx5RLn//EHJ3z33dltw5hXEfqUOHDuHh4UHNmjVvWnb9+vUAtGuXc8+zkJAQihcvzv79+zl/PuuX5OjoaE6ePEnp0qVp3Lhxjjrt22e9h3Xt2rWWYxs3bsRoNBISEoK3d85XG1mrk1+FekTq3LlztGzZkiJ5LLK8kYeHB/fddx8bN268jT2T3Cx582GaBJbi299iGL3I+nRcgwBfAKL+sZVBzXLFWTz4YUoX9yAlPYOXv9rJ6n1n89WmwQCNq2X9Jr7pcFyeZfN6P19xLzfq/r0HlrnchaSbL94VKYiES/FMHPoScWdP4+LiyvMvv23ZVNOa6rXv4VRMNNGH99Oy/dM5zh87ehCj0Yh7EQ+q3jCKdHjfLj4aPZj0tDS8fUrw5ntTqFm3Ya5tlSxdltJl/bl44Tx/HTlIw8Y5t1DINBo5H5v1yiW/8hXze9tyFzl16hRXrlyhdu3aLFiwgB9//JGYmBiKFClC48aNGThwIA0bZv05jI+P5+LFi7i7uxMYGJjjWq6urgQGBrJ3716OHj2Kv78/kZGRANSsWdNqADQfP3nyJCkpKXh5eVnq1KplfaS1Ro2sNbZHjuT9cntrCnWQKl26NOfOnStwvVOnTlG0aNHb0CO5maOxV3iwZhk6NKrAB+GHuHbDa1cqlvLiqcZZo0LhO7P+MS5R1J15r4RQurgHicnpPP/5tjzXWd2oVvniFPPM+qP8Z0ze9fJ6P1/TmmUso1L5eUegSEFlXL/Ox++9SdzZ07i5uTNo2ASaPNQqzzrBzR9hw8of+WPzOrr3exXv4iWynV/78xJLuSL/WOMUd+4Mn4x7m/S0NEqV8cvxupfcNG31KBGLvmX9Lz/yaKceOXZgX/fLMtLTUilazJv69wXl99blNipsM3vm9VFHjx7lgw8+4P777+fBBx/kyJEjrF27lo0bN/L+++/z1FNPEReX9ctvmTJlcl2b5OfnB2Apa/7s7+9vtbyHhwc+Pj4kJiYSHx9PQEDATevc2EZBFOogdf/99/Pzzz+zbNkyOnXqlK863333HYcOHbrpY5dye3y5NppnQqpQvqQXX7zwAG+F7eLC3/tF1avowxcvBFHM041tkfGs2JM1VTEq9B4qlS6KMdPES7P/KFCIgv+NcF1Juc7xPJ7WE3G0iMVziYnK+o2396B3bhqiIOtVLrXq30vkwayXGb86/H1KlCyd9dLiJfPYun4lrm5uPPH3k35mX30ykeRrVyni4cFb46bmK0QBPN7lWTatWc7li3FMGfMmLw0ZR+myWT989v+5ncVzpgPwZLfeeS5ulzvHXtNy33//PYsWLSpQnW7dutG9e/dsx/75xN706dOpUiXrz15mZiZffvklH3/8MSNGjODee+8l+e8tN8wPkFljfrrPXPbatWv5rmMua67r5eVltbz5WuZyBVGog9Qrr7zC2rVrGT58OBERETzyyCPUqlULf39/vLy8MBgMpKSkcOHCBaKioli9ejWbN2/Gy8tLC80dJObCNV79Zief/l8T2txTjh0T2nMs7ipuLi7ULJ/1BNCe45d5YeZ2TCbw8/Ggy9/rkFLSMxjcoQ6DO+T++On3W0+wcNuJbMfK+Wb9xThzqeB/AUTulIzr11m5bAGQNV2xac3PlvfaWXPvAyE81b0PBoOBAW+9x4R3BnD0wB7e6P00laoGcin+AomXL2IwGHjxjdGWTTUh6z195le9FPHw5NtpH+bZt6e69+HeB0IA8C5egrfGTuU/owZz9MBu3urTkQqVq5Gelsb52KwnYpu2bE+HLs/d0vdDCp8LFy5w8GDeb4SwVudGr7/+Ol27dsXb2zvbPlEuLi4MGDCA3bt3s379ehYsWGBZF5WfMGje/si8ULwgATK/dXLbYikvhTpIVa1alW+//ZY333yTLVu2sHXr1jzLm0wmKlSowOTJky3znXLnrdgdy9HYdfRvU4Pmdf0I9PMm9bqRnX9dZNkfpwnbFIPx720HmlQvjbtr1nCut6c7QTXyfk3LpiM5h11Le2f95hF7OcXOdyJiP6eOR3PtataTokajkchDe/Ms71fhf6+E8StfkYmfzyP8+6/ZvX0Tp2L+ooinJw2bNOXxLs9R794m2eoe2b/b8vXVK4k3bSsxIfsocLWadflg1kJWLJnPrm0bOXfmFB4eHtRt2JhWj3Wkacv2+bpnuTPsNbVXtmxZ6tevX+A6N3Jzc6Ny5dwf1GnTpg3r169n//79ltmm1NTc30Rh3mjWvGTHvLVRfuqYR6DMdVJSrP+cMF8rtxGrvBTqIAXQsGFDVq1aRUREBOvXrycyMtKys7mLiwvFihXD39+f2rVr06JFC9q2bZttky9xjL/OX2Xod3tuWm7F7lgqvrTsltoau2R/vp/uy8u2qPhb7otIbqrVrEvYL7/bXL94CV+eHfBmtp3Ic9Ohcy86dO5103J5tufjyzN9BvFMn0G3dB25/ew1tde9e/cc03S3g3lvp5SUFMuapfj4eEwmk9V7Ma9bMq9jMtexNhoGWSHqypUrGAwGS9C7WZ0b2yiIQh+kIGtI7umnn+bpp3M+sSIiIiKFx8SJEzl79iyvv/661e0Pzp7NeiK7XLly+Pr64u/vz/nz5zlx4gRVq1bNVtZoNHLs2DEAateune1zVJT1V4FFRUVhMpmoXLmyZRTrZnWio6OzlSuIQr2PlIiIiOTNYLDPh70cPHiQNWvWsGLFCqvnw8PDAWjZsiWAZQPM1atX5yi7ZcsWkpKSqFOnDuXKlQOgSpUqVK1albi4OPbs2ZOjjnkfSfP1AZo3b46LiwubN2+2uqB81aqs91y2anXzB0BupCAlIiLixFxcDHb5sJdevbKmlWfPns327dstx41GIx9++CF//PEHVatWtcwy9erVC1dXV2bOnJktGMXGxjJ+/HgABgwYkK2N55/Pekp11KhRxMfHW45v27aNuXPn4u7uTt++fS3H/f39adu2LQkJCYwZM4b09P9tOGtut0aNGrRpU/BXHjnF1J6IiIhYV9j2kXr88cf5/fff+f777/m///s/7r33Xvz9/Tlw4ABnzpyhbNmyTJ8+3bKeuU6dOgwePJgpU6bQs2dPgoKC8PT0ZMeOHSQnJ9O1a1c6dOiQrY0ePXrw22+/sWHDBtq3b09wcDBJSUns3LkTk8nE5MmTs71nD7JeJ3Po0CF++ukndu7cSYMGDYiJiSEyMhIfHx8+/vhjm9abGUy2POt3F9NiZJGclr3T2tFdECmUggJL3LzQLao/IueUmC0OTsz5ipZb8csvv1j2dkxPT6d8+fK0bt2a/v37Z9sWwWzt2rXMmTOHgwcPYjAYqFatGj169KBTp05WN+u8fv068+bNY+nSpZw4cQJvb2/q1q3LgAEDCAqyvlnspUuXmD59OuvWrSMuLg4/Pz+CgoIYNGgQAQEBNt2nglQBKUiJ5KQgJWLdnQhS94xcY5frHJjQ1i7Xudtoak9ERMSJFbapvbuNFpuLiIiI2EgjUiIiIk7MXhtyim0UpERERJyYgpRjaWpPRERExEYakRIREXFiGpByLAUpERERJ6apPcfS1J6IiIiIjTQiJSIi4sQ0IOVYClIiIiJOTFN7jqUgJSIi4sSUoxxLa6REREREbKQRKRERESemqT3HUpASERFxYspRjqWpPREREREbaURKRETEiWlqz7EUpERERJyYcpRjaWpPRERExEYakRIREXFimtpzLAUpERERJ6Yc5Via2hMRERGxkUakREREnJim9hxLQUpERMSJKUc5loKUiIiIE9OIlGNpjZSIiIiIjTQiJSIi4sQ0IuVYClIiIiJOTDnKsTS1JyIiImIjjUiJiIg4MU3tOZaClIiIiBNTjnIsTe2JiIiI2EgjUiIiIk5MU3uOpSAlIiLixJSjHEtTeyIiIiI20oiUiIiIE3PRkJRDKUiJiIg4MeUox1KQEhERcWJabO5YWiMlIiIiYiONSImIiDgxFw1IOZSClIiIiBPT1J5jaWpPRERExEYakRIREXFiGpByLAUpERERJ2ZAScqRNLUnIiIiYqN8jUj16tXrlhsyGAzMmzfvlq8jIiIi/6On9hwrX0Hqzz//zPWc+WkBk8mU5zk9VSAiImJ/+vnqWPkKUq+88orV41u3bmXXrl34+vrSoUMH6tati4+PD6mpqURHRxMREcG5c+do27Ytbdu2tWvHRURExLm89tprrFq1ivfff5/Q0NAc57du3cqXX37JkSNHSE1NJTAwkO7du9OlSxergTEjI4MlS5awcOFCjh8/jpubGw0aNKB///4EBwdb7UNiYiKzZ89mzZo1xMbG4uPjQ7NmzRg0aBABAQEFviebg9SBAweYMWMGQUFBTJs2DR8fnxxlBg0axOuvv87atWvp2bNngTsnIiIieXOWAanFixezatWqXM8vWLCA9957D3d3d4KDg3F3d2f79u2MHDmSXbt28f7772crbzKZGDp0KBEREZQoUYKQkBASEhLYunUrW7ZsYeLEiXTu3DlbnYSEBHr16kV0dDSVK1emZcuWxMTEsGzZMtasWcP8+fOpU6dOge7L5sXmn332GQaDgSlTplgNUQCenp5MnjwZd3d3pk+fbmtTIiIikgsXg8EuH7dTTEwMkyZNyvP8+PHj8fb2ZtGiRXz11VfMmDGDFStWULlyZZYuXcrKlSuz1VmyZAkRERHUrVuX1atXM23aNObNm8fXX39NkSJFGDt2LGfPns1WZ9KkSURHRxMaGsrKlSv59NNPWb58OW+88QZXr17lnXfeITMzs0D3ZnOQ2rVrFzVr1qRs2bJ5litZsiQ1a9bk4MGDtjYlIiIiuTAY7PNxu6Snp/PWW2/h4uJCvXr1rJaZPXs2RqORfv36ZStToUIFRo8ebSnzTzNnzgRg5MiR+Pr6Wo43bdqU3r17k5aWRlhYmOX46dOniYiIwMfHhxEjRuDq6mo5N3DgQBo1akRkZCSbNm0q0P3d0vYH165dy1e5S5cuUaRIkVtpSkRERJzQ1KlTOXjwIKNHj6Z8+fJWy6xfvx6Adu3a5TgXEhJC8eLF2b9/P+fPnwcgOjqakydPUrp0aRo3bpyjTvv27QFYu3at5djGjRsxGo2EhITg7e2drzr5YXOQql69OidPnmT79u15lluxYgWxsbHUrVvX1qZEREQkFwaDwS4ft8PWrVv55ptvePzxx3n66aetlomPj+fixYu4u7sTGBiY47yrq6vl+NGjRwGIjIwEoGbNmlb7bj5+8uRJUlJSstWpVauW1X7UqFEDgCNHjhTkFm3f2bxr167s2bOH1157jREjRtChQwfc3d0t51NTU1m8eDFTpkzBYDDw7LPP2tqUiIiI5MJeGej7779n0aJFBarTrVs3unfvbvXcpUuXGDJkCOXKleO9997L9RpxcXEAlClTBhcX6+M7fn5+2cqaP/v7+1st7+HhgY+PD4mJicTHxxMQEHDTOje2kV82B6nQ0FDWrVvH2rVrGTZsGGPGjKFy5coULVqUa9eucfz4cTIyMjCZTDz77LO0adPG1qZERETkNrtw4UKB1zNfuHAh13MjRozg4sWLzJkzJ9eH0gCSk5OBrAfUcuPh4ZGtrHlpUX7qmMua63p5eVktb76WuVx+2RykDAYDn3zyCbNnz+brr78mKSnJMmxmVrp0aV5//XW6detmazMiIiKSB3s9cVe2bFnq169f4DrWzJ8/n3Xr1vHiiy/mup+TmXkUKj/Ti+bNv80LxQsyJZnfOtY2GM/LLb202N3dnZdeeokXX3yRnTt3EhMTw5UrV/D19aVatWo0adIk12E6ERERuXX2Wt3UvXv3XKfpCiIqKooPP/yQ+vXr8/rrr9+0fLFixYCsJUG5SUtLA6Bo0aIFrmMegTLXMa+ZupH5WrmNWOXmloKU5SJubjz44IM8+OCD9riciIiIOKmPPvqI1NRUPD09effdd7OdM08dLlq0iK1bt/LAAw9YnpaLj4/P9ZVy5nVL5nVM5nVOuU0tpqWlceXKFQwGg2XU7GZ1bmwjv+wSpBITE9m2bRvHjh0jKSmJoUOHkpaWxt69ewkKCrJHEyIiImJFYXvXnnmN0Z9//pnru3p3797N7t27cXNz45lnnsHf35/z589z4sQJqlatmq2s0Wjk2LFjANSuXTvb56ioKKvXj4qKwmQyWdZu56dOdHR0tnL5dUtBymQy8dlnn/HNN99kG14bOnQop0+fpnfv3jRq1IjPP/+cUqVK3UpTIiIiYoVL4cpR2TbBvNHLL7/M2rVrc7xrr0WLFixatIjVq1fTv3//bHW2bNlCUlISderUoVy5cgBUqVKFqlWrcvz4cfbs2UOjRo2y1THvgt6yZUvLsebNm+Pi4sLmzZtJTk62BCwz8+trWrVqVaD7vaUFTEOGDOGLL74gJSUFX1/fbPOKCQkJmEwm9uzZw3PPPZfrnKSIiIjc3Xr16oWrqyszZ85kz549luOxsbGMHz8egAEDBmSr8/zzzwMwatQo4uPjLce3bdvG3LlzcXd3p2/fvpbj/v7+tG3bloSEBMaMGUN6errlnLndGjVqFHiXAZtHpFavXs3y5cspXbo0kydP5uGHH6Znz57s3r0bgMaNGzN//nxef/11jh07xty5c3N8E0REROTWFLapPVvUqVOHwYMHM2XKFHr27ElQUBCenp7s2LGD5ORkunbtSocOHbLV6dGjB7/99hsbNmygffv2BAcHk5SUxM6dOzGZTEyePDnHTuojR47k0KFD/PTTT+zcuZMGDRoQExNDZGQkPj4+fPzxxwX+fto8IrVw4ULLS4sffvhhq2UaN27MJ598gslkyvONzyIiImKbwv6uvfzq378/06dPp3Hjxuzbt48//viD6tWrM2nSJMaNG5ejvIuLC9OmTWPYsGFUqFCBzZs389dffxESEsLcuXPp2LFjjjp+fn4sWrSI5557DoPBwLp167h27RqdOnVi6dKlBV4fBWAwFXTDhL8FBwdTrFgx1q1bZzlmHpE6fPhwtrJt2rTh8uXL7Nq1y5amCpWKLy1zdBdECp1l77R2dBdECqWgwBK3vY3nv9tnl+vM7dnQLte529g8IpWcnJztbct5KVWqFBkZGbY2JSIiIlIo2bxGqkyZMpw4cSLXPR/Mrl+/zvHjxylTpoytTYmIiEguCttTe3cbm0ekgoKCSE5O5rvvvsuz3LfffktSUhJNmjSxtSkRERHJhcFgsMuH2MbmINWnTx9cXFz44IMPmDt3LpcvX852/uLFi3zyySd8/PHHuLi48Oyzz95yZ0VEREQKE5un9urUqcPw4cOZMGEC77//Pu+//77lXNOmTUlISACyNu18/fXXadhQi9hERETsTWNJjnVLO5v36tWLgIAAPv74Y44cOWI5bh6dqlKlCq+//nqOvR9ERETEPlw0LedQNgepzMxMXFxcaN68Oc2bN+fMmTNERUWRlJSEl5cX1apVo3r16vbsq4iIiEihYnOQ6tu3L2XLlmXUqFH4+PhQsWJFKlasaM++iYiIyE1oQMqxbA5SBw4cwMvLCx8fH3v2R0RERApAT9w5ls1P7RmNRkqXLm3PvoiIiIg4FZuDVJs2bYiMjOTPP/+0Z39ERESkAP4t79pzVjZP7b377rvExsbSt29fHnvsMRo3bkzZsmXx8PDItU7Tpk1tbU5ERESs0FN7jmVzkGrWrJnl6/DwcMLDw/MsbzAYOHTokK3NiYiIiBXKUY5lc5AymUy3tbyIiIhIYWdzkPrnBpwiIiLiGHpqz7FuaWfzu9Ffn3VydBdECp2SD7zi6C6IFEopu6fd9jZsfmpM7MJuQer48eMcP36cK1euULp0aWrUqIG/v7+9Li8iIiJS6NxykFqxYgXTpk0jJiYmx7lGjRrx5ptv8sADD9xqMyIiImKFpvYc65ZGBCdOnMhbb73FsWPHMJlMeHt74+fnR9GiRTGZTOzevZvnn3+esLAwe/VXRERE/sHFYJ8PsY3NI1K//vorYWFhuLm50a9fP7p370758uUt50+dOsV3333Ht99+y+TJk7n33ntp2LChXTotIiIiUhjYPCIVFhaGwWBgzJgxvPHGG9lCFEBAQABDhw5l+PDhGI1Gvvnmm1vurIiIiGSnESnHsjlIHT16FH9/f7p27ZpnuV69elGmTBm9SkZEROQ2MBgMdvkQ29gcpNLT0ylTpsxNyxkMBsqXL8+VK1dsbUpERESkULI5SNWqVYuoqCguX76cZ7nU1FSOHTtGzZo1bW1KREREcqGpPceyOUgNHDiQtLQ03n77bVJTU3MtN2HCBJKTk+nTp4+tTYmIiEguDAb7fIhtbH5qr1y5cjz77LPMmzePDh068Mwzz9CwYUNKlChBcnIyUVFRLF26lAMHDlCzZk2Sk5P54YcfclynS5cut3QDIiIidzMXpSCHMphsfJtwnTp1LIvTTCZTrgvV8joHcPjwYVuad5jUDEf3QKTw0StiRKy7E6+IGbYi0i7Xmdyhll2uc7exeUSqQoUK9uyHiIiI2EDv2nMsm4PUunXr7NkPERERsYFm9hzLYUF2woQJ9O7d21HNi4iIiNyyW35psa0OHTrE7t27HdW8iIjIv4IWmzuWw4KUiIiI3DrlKMfSGjURERERG2lESkRExIlpV3LHUpASERFxYloj5VgKUiIiIk5MOcqxtEZKRERExEYakRIREXFiWiPlWApSIiIiTsyAkpQjaWpPRERExEYakRIREXFimtpzLAUpERERJ6Yg5Vg2T+1duHDBnv0QERERcTo2B6lWrVoxcOBAVq9ezfXr1wtcv0yZMpQvX97W5kVERAQwGAx2+RDb2Dy1ZzQa2bBhAxs3bqREiRI8+eSThIaGUrdu3XzV//TTT21tWkRERP6mqT3HsjlIrV+/nh9//JHw8HBiYmIICwtj3rx51KlTh9DQUJ544glKlixpz76KiIiIk/jll1+YP38+Bw8exGQyERAQQIcOHejTpw+enp45ym/dupUvv/ySI0eOkJqaSmBgIN27d6dLly5WR8wyMjJYsmQJCxcu5Pjx47i5udGgQQP69+9PcHCw1T4lJiYye/Zs1qxZQ2xsLD4+PjRr1oxBgwYREBBg030aTCaTyaaa/7B3716WLl3KypUrSUxMxGAw4ObmRqtWrQgNDaV58+a4uPw7dlpIzXB0D0QKn5IPvOLoLogUSim7p932Nj7+7ZhdrvNm80C7XAfgs88+Y9q0abi6utK4cWO8vb3Zt28f8fHx1KxZk/nz51OiRAlL+QULFvDee+/h7u5OcHAw7u7ubN++nZSUFEJDQ3n//fezXd9kMvH2228TERFBiRIlCAoKIiEhgT///BOTycTEiRPp3LlztjoJCQn06tWL6OhoKleuTN26dYmJiSEyMhJvb2/mz59PnTp1CnyvdglSZunp6axbt47w8HC2bt1KWloaBoOB0qVL07FjRzp16kT16tXt1ZxDKEiJ5KQgJWLdnQhSn2yKsct1Bj9czS7X2blzJ7169cLHx4ewsDBLOElOTua1115j06ZN9OzZkzFjxgAQExPD448/jpeXF2FhYdSrVw+A2NhYevfuzcmTJ/nvf//Lo48+amnjhx9+YMSIEdStW5c5c+bg6+sLwLZt2xgwYAAAq1atyrYWe8iQIYSHhxMaGsqECRNwdXUFYMaMGUydOpVatWoRHh5e4IEfuw4TFSlShEcffZQvvviCHTt2MGzYMLy8vLh48SJfffUVTzzxBL169eLXX3+1Z7MiIiJ3LReDfT7sZdmyZQC8+OKL2UZ4ihYtymuvvQbAhg0bLMdnz56N0WikX79+lhAFUKFCBUaPHm0p808zZ84EYOTIkZYQBdC0aVN69+5NWloaYWFhluOnT58mIiICHx8fRowYYQlRAAMHDqRRo0ZERkayadOmAt+v3efbTp8+zYwZM+jZsycffPABycnJmEwmateuTenSpfnzzz959dVXefHFF0lOTrZ38yIiIuJAY8eOZcWKFXTv3j3HOaPRCJAtyKxfvx6Adu3a5SgfEhJC8eLF2b9/P+fPnwcgOjqakydPUrp0aRo3bpyjTvv27QFYu3at5djGjRsxGo2EhITg7e2drzr5ZZcNOZOSkvjll18IDw9n165dQNb8pflpvi5dulCnTh2MRiNr165l3LhxbN68mYkTJzJx4kR7dEFEROSuVNh2LnBzc7O6jOfs2bN88MEHAISGhgIQHx/PxYsXcXd3JzAw5xotV1dXAgMD2bt3L0ePHsXf35/IyEgAatasaXURuvn4yZMnSUlJwcvLy1KnVq1aVvtco0YNAI4cOVLw+y1wjb8ZjUY2btxIeHg4GzZsID09HZPJhIuLC02bNqVz58488sgjFClSxFLH1dWVdu3aUbRoUV544QV+/fVXBSkREZFb4GKnlxZ///33LFq0qEB1unXrZnXk6Z8mT57M3r172bt3LwaDgb59+zJw4EAA4uLigKy9JXNbm+Tn55etrPmzv7+/1fIeHh74+PiQmJhIfHw8AQEBN61zYxsFYXOQatasGQkJCZjXqgcEBNCpUydCQ0MpV65cnnXNqfOfQ3siIiLiOBcuXODgwYMFrnMzS5Ys4cqVK0DWWuoLFy4QHx+Pn5+fZYmPte0QzDw8PAAsZa9du5bvOuay5rpeXl5Wy5uvZcuSI5uD1OXLl/H09KRdu3Z07tw51z0brElLS6Nbt240aNDA1uZFREQE+03tlS1blvr16xe4zs389NNPlCxZksjISKZMmcLy5cvZvXs3y5cvt4xC5WdndfPAjXkQpiC7see3ji0bGdgcpMaOHcvjjz9uddHWzVSrVo1x48bZ2rSIiIj8zV5P3HXv3v2m03S2MG9B0LBhQ7788ks6d+5MZGQkixYtomnTpgCkpqbmWj8tLQ3IeuoPoFixYvmuYx6BMtdJSUmxWt58rdxGrPJi81N7zzzzjE0hSkRERO5ORYoU4bHHHgPg0KFDljVL8fHxuY4GmdctmdcxmevkNq2YlpbGlStXMBgMlhGzm9W5sY2C+HdsNy4iInKXcjEY7PJhL59++imDBw/m3LlzVs+bH0LLyMjA19cXf39/0tPTOXHiRI6yRqORY8eydm6vXbt2ts9RUVFWrx8VFWV5JY15FOtmdaKjo7OVKwgFKRERESdmMNjnw162bNnCL7/8ws8//2z1/MaNGwEs66RbtGgBwOrVq61eKykpiTp16lgeZKtSpQpVq1YlLi6OPXv25KizcuVKAFq2bGk5Zn5V3ebNm60uKF+1ahUArVq1yudd/o+ClIiIiNhNr169AJg2bRr79u2zHL9+/TofffQRv//+O6VLl7a8C69Xr164uroyc+bMbMEoNjaW8ePHA1he+2L2/PPPAzBq1Cji4+Mtx7dt28bcuXNxd3enb9++luP+/v60bduWhIQExowZQ3p6uuWcud0aNWrQpk2bAt+vXd+1dzfQu/ZEctK79kSsuxPv2vvq95N2uU6/oMp2uQ7A6NGjWbhwIS4uLtx33334+Phw+PBhzp07h6+vL7NmzeLee++1lJ81axZTpkzB1dWVoKAgPD092bFjB8nJyXTt2pUJEyZku35mZiYvvfQSGzZswNvbm+DgYJKSkti5cycmk4nJkyfTsWPHbHXi4uLo2bMnp06dokKFCjRo0MDy0mIfHx/mzZtn09SeglQBKUiJ5KQgJWLdnQhSX/9hnyDV9wH7BSmAFStWsGDBAg4ePEh6ejoVKlSgZcuW9OvXz+rGmGvXrmXOnDkcPHgQg8FAtWrV6NGjB506dbK6Wef169eZN28eS5cu5cSJE3h7e1O3bl0GDBhAUFCQ1T5dunSJ6dOns27dOuLi4vDz8yMoKIhBgwYREBBg030qSBWQgpRITgpSItbdiSA1x05B6v/sHKTuFlojJSIiImIju7y0WERERByjIDt8i/0pSImIiDgxxSjH0tSeiIiIiI00IiUiIuLE7LkruRScgpSIiIgTU4xyLE3tiYiIiNhII1IiIiJOTDN7jqUgJSIi4sS0/YFjaWpPRERExEYakRIREXFiGhFxLAUpERERJ6apPcdSkBIREXFiilGOpRFBERERERtpREpERMSJaWrPsRSkREREnJimlhxL338RERERG2lESkRExIlpas+xFKREREScmGKUY2lqT0RERMRGGpESERFxYprZcywFKRERESfmosk9h9LUnoiIiIiNNCIlIiLixDS151gKUiIiIk7MoKk9h1KQEhERcWIakXIsrZESERERsZFGpERERJyYntpzLAUpERERJ6apPcfS1J6IiIiIjTQiJSIi4sQ0IuVYClIiIiJOTNsfOJam9kRERERspBEpERERJ+aiASmHUpASERFxYpracywFKXGIHdu38f2C+ezbs4eEhARKlvTlgeAHebH/SwRWr56va/x36hS+nj2LmbO/4cGmIbmWy8zM5MdlS1ge/iPRUVGkpCRTvkJFWrVuzQsvDsSnRAl73ZZIvrR4oBYvPdOcoIbVKO1bjIsJ19jw+1E++GoVR2PO5yjv5ubCxa0fU8Q993+yE5KSKd98CACVy5fi6Ipx+e7PhBkrmDhzRbZj9aqXZ9iLj9K8SU18i3txLv4KKzcf5MPZq4i9kJjva4v82ylIyR33yccf8c1XXwJQtmxZAgMDOX48hhURy1m7ZjX/nfYFTUMeyvMaGzesZ+6cr2/aVnJyMq+/8hK/79gOQJWqVSlJSU6dPMm333zNmtWrmDP3O/zLlbv1GxPJh/GvPcXbfdoBcPZCIkdizlGrij89Hg+iY5tGdBk8i3U7jmSrUzewPEXc3biWksbeI6etXjcpOdXydVr6dbbu/ivPfviX9qF65bIA/HUyLtu5h+6rzvLpg/DyLMKFy0kciD5LrSp+DOjWnK7tG/NY/0/ZF3mmwPcut4ee2nMsBSm5o5Yt/YFvvvoSNzd3Rr83jqc6dsJgMJCYkMCI4UPZtHEDw4e9w88rf6Vo0aJWr7Fq5QpGvjuUjIyMm7Y3cfx7/L5jO2X9/Jj6389p0LAhAFGRR3lz8GucPHGc8WNHM+2LWXa9TxFrnn/6Qd7u04706xkMGr+Aect3AFDSpyizxz9Ph+b38PXE56n3xHskp6Zb6jWoVRGAzbui6fjKFzdt5/zFJNr0nZrreU8PdzbPeweAb5Zt5ftfdlrOlfQpyqKp/fHyLMJH36zmvc8jMBoz8S7qwayxz9LpkftY8NELNAqdwPUMo03fB7EvTe05lp7akzsmLS2NqR/9B4Ch7w7n6U6hGP7+VaqEry/vf/ARxYoV49LFi2xcvy5H/StXrjBp/FiGvPUG6enpOc7faP++fUT8FI6rqytfzJxtCVEANWvVZtSYsQBs3vQb58/nnE4RsSePIm5MGtwJgLf/84MlRAFcvpJMnxFzuHI1Bf/SPjzRskG2ug3/DlIHo2Lt0pcpQ7pQv0YFjsac460PF2c793KPlpQqUYwd+2IY9elPGI2ZAFxNTqP3u3OIOR1PYEBZej0ZZJe+yK1zMdjnQ2yjESm5YzZuWEdiYgKVq1Slc9dncpwvXrw4w4aPIiEhgarVqmU7t3fPbl4bNJCEhAQ8PT15d+Roxowcnmd7y8OXAfDk0x2pWat2jvMPBAXzymuD8fYujquLfqeQ2+vx5g0o7VuMqBNxfLVkS47zV66m8uaHP1C6RDEij2cP9uYRqYN/nb3lfrQMqkXf0IfIzMzk5XHfkZJ6Pdv5554KBmDOj1tz1L2eYWTOj9sY+8qTdGvfhDnLtt1yf0ScnYKU3DHbt2b9w9yqdWtcXV2tlnmqYyerx4/HxJCQkECzh5sz9N2RVK5S5aZBatu2rPbaPNLO6nmDwcCLA17Kb/dFbknrB+sAELFhH5mZJqtl5v9jlOqfGthpRMrFxcB/3u4CwHc//8HWPceynS9XxocqFUoDsHX3sRz1Abb9Xadpo0Dc3FzIyMi8pT7JrdPUnmMpSMkdExl5FIDq1WtiMplY++saNq5fx/nz5yhRogRNH2rGk091xN3dPUfdGjVrMifsO+67v3G+2kpJSeH0qZNZ7dWowbVrV/l5+U/8vmM7V65coXyFCrRr/xgPNXvYfjcokocGNSsAcOjvUaWnW9/L4y0aUNHfl8uJyazZdpj5ETtyBJPyZUtQtmRxMjKMJCWnMqRfex5oUJWinu4cP3OR8HV7Wb3lUL760Df0Ie6pWYHUtOu8N215jvPVA7IWn2dmZnL8zEWr1zgRm3Xc08OdgHKliDkdn79vgNw2WmzuWApScsecjc36bdrN3Y2+vZ9l1587s51fvWol382by7TpsyhXvny2c/Xvyb5m5GbOnztLZmbm31+fo1/v5zh7Nvtv8z8uXUL7xzowYdIHFClSpKC3I1IgAeVLAZCRYWTNV4Npdn+NbOc7t7ufQT1aEvraF5w+n2A5bh6NMplg56LheHlm/7PaN/QhVvx2gN7vfsPV5LRc23dxMfDW/z0CwPyI3zkTl5CjTNlSxQG4ci2V9OvWH+a4lHjN8nUZ32IKUnLX08IQuWOuXcv6B/g/H7zPwQP7GTJsOOs3bWP7zj18+vkMKlSsSFRkJK++PIDr+VhMnp+2AN4c/CoGg4FPPvuc7Tv3sHHLdt4dMRpPT09W/bKCDydPvKW2RPKjeDFPAD54uzON61Xm7Q9/IKD1MEo1fYPQ12Zw/Ew8DWpVZMmnA3F3+9/Ut3mhubu7Kyt+O0BIzw/wDR5MlTbv8vqkhVy5mkKH5vcQ9kHfPNvv2LoRVSuWwWjMZOrcX62W8fLMGg1OTcv9idiUtP+tqbox1IljGOz0IbZRkJI7Ji0ta5+by5cuMfk/H9Prud6UKlUKLy8vWrRsxRczZ+Pm5k5k5FF+Cv/xFtv632/m19PT+erbMFq1fgQvLy98fUvSvWcv3h05GoAlixdxPMb6ehARe/HyyAopfqWK0/vdOXy+YAPxl6+SknqdXzYd4KlB00m/nkHDWpV49slgS739UWeYueg3xn0RwbNDv2b34VOkpWcQdymJWYs30enVL8jMzOTRZvVp91C9XNsf1LMlAD+t38tfJy9YLZNpzFq7ZTJZX8OVo3w+y8nt5WIw2OVDbKMgJXeMp2fWb+S1a9ehdZtHcpyvWi2Qxzo8DsD69Wvt0hbA0x1DqVChYo4y5uOZmZls3LD+ltoTuRnzvlB7j55m+YZ9Oc5HnYhj0co/AXii5f+26li1+RCD31/E+7NWWr3u1j3HWLcja/3hU63utVqmop8vD96b9STsdz//nmsfr6Zk/QLi6ZFznaKZ1z/Opd7wxJ/IjcLDw3nuued44IEHuOeee2jRogVDhw7lr7+sbxi7detW+vTpQ9OmTbnvvvvo3LkzixcvzjXcZ2RksHDhQkJDQ7n//vsJCgqiX79+7Nhh/cENgMTERKZMmcKjjz5Kw4YNadasGcOGDePUqVM23aOClNwxxX18AKhdp26uZWrUrAnAGRv/QFvaKu5j+Tq39gwGA9VrZK1TOX2L7YncTGJSCgD7jlrfmRzgUHTWOr5qlUoX6Nrm3c6rVrRe78lWDXFxcSExKYXVWw7nep2LCVcB8CnmiZub9R8PpX29LV/HXUoqUD/l9iiMU3smk4m33nqLIUOGsGvXLqpXr07z5s1xdXXlxx9/JDQ0lM2bN2ers2DBAvr06cMff/xBvXr1CA4O5q+//mLkyJEMH57zKW2TycTQoUMZPXo0p0+fJiQkhFq1arF161Z69+7NkiVLctRJSEigZ8+ezJo1C6PRSMuWLSlZsiTLli2jY8eOHDlyJEedmyn0i80nTJhwS/VHjhxpp57IraoWWJ3z587luZmmq2vWH0n3W1z8XaFiRTw9PUlNTSX9el7tZa1F0WJzud2OxJyjUrmSeBTJ/Z/djL83v0xLz75GycXFgMFgsGyOeSOXv3dTzG2ncfMI1/IN+3JdRA5w5Ng5AFxdXXJ9Iq9KhaxF8ymp6Zw+fznXa8kdVAhn5X766SciIiIoW7Yss2fPpk6drO0/jEYjn376KTNmzGDIkCGsWbOGYsWKERMTw/jx4/H29iYsLIx69bKmqWNjY+nduzdLly6lRYsWPProo5Y2lixZQkREBHXr1mXOnDn4+voCsG3bNgYMGMDYsWMJCQmh/D8eXpo0aRLR0dGEhoYyYcIEy8+AGTNmMHXqVN555x3Cw8NxKcDegoU+SC1ZsoTU1P+9Qyq/c/eQNeKgIFV4NGx4L9u3buHAgZzTGmYxf69VCgiofEttubq6Uq/+Pez6cyf79+2la7fuVssdPx4DQKVbbE/kZn7ff5xHmtalSf0quZapVdUfIFuAOfLzWALKleTdqT/y6bycO/4DNKoTkFX27yD0TwaDgeCGWdN6N77D70YJSSlEnYijZhU/ghtWsxqkHrw3EIA/DpzIdT8skR9++AGAt956yxKiIOvf5sGDB7N27VqioqLYsmUL7dq1Y/bs2RiNRvr162cJUQAVKlRg9OjRvPDCC8yePTtbkJo5cyaQNWBiDlEATZs2pXfv3syaNYuwsDCGDMl6mffp06eJiIjAx8eHESNGZNvPcODAgaxfv549e/awadMmWrRoke97LfRTe8uWLaNatWqYTCYqVarEK6+8ku+PQYMGObr78g8dHn8CyJpGW/vrmhznL168yMoVEQA80q79rbf3xJMArF65knNnc+4IvXnTRo7HxODi4kLrR9recnsieVn49/vsAgPK8lSrhjnOly3pTbdHs/ZJW/brHsvxw8fO4eLiQvcOD+DqmvOf7Cb1q9Digawp8cWr/sxxvm5gObyLegCwY1/MTfu5ZPUuAPp1zvnicHc3V3o//SAAYT9tv+m15M4w2Ol/9uTj40P16tVp0qRJzv4aDFT7++0V5tdzrV+ftU61XbucGyiHhIRQvHhx9u/fbykfHR3NyZMnKV26NI0b59xfsH37rJ8ha9f+b73txo0bMRqNhISE4O3tna86+VHoR6SqVq1KWFgYPXr04NSpU9xzzz20bNnS0d0SG1QLrE5o564sXbKYMSPfxdXVlZatWgMQf+ECQ995k2vXrlGrVm0esUOw6dgxlAXz5/FXdBQvD3yR/0z5xLIm6tDBA4x7L+upvc5du+Hv73/L7YnkJfL4eb5euoW+oQ8xc+yzZBjnsuK3AwD4ly7Ot+/3wcfbi32Rp/lx7R5LvSlz1tAupC731Q3gi9E9eevDH0i6ljVK/3Djmnz7/v/h4uLCdxG/s+vQyRztNqqbNVqVmJTCsVM33/Np+oIN9O/2MM3ur8GUIV0Y9vEyrmcY8S7qwcz3ehEYUJZjpy7w/S9/2OG7IvZQGB+4+/zzz3M9ZzQaOXjwIADly5cnPj6eixcv4u7uTmBgYI7yrq6uBAYGsnfvXo4ePYq/vz+RkZEA1KxZ0/LO1n8yHz958iQpKSl4eXlZ6tSqVctqv2r8/fOhoOukCn2QAihdujT//e9/eeaZZxg3bhwhISFa0+Kkhg4fyYX4C2zauIHXX3mJcuXKU7JUKaKjIrl+/ToVKlTkwylTb3mNFGSts/r08y946cV+/BUdReeOT1CtWiAYDBz7KxqA4Aeb8ubbQ2+5LZH8eOvDHyhXpgQdmt/Dkv8O5NTZS1y4fJX6NcrjUcSdE7EXeW7oN9nWOm3+M5p3PlrCB2+G8txTD9K57f1EnjhPCW8vqlUqA8Avmw7w8vjvrLZZwc8XgFPn8ree6cLlq7wwKowFH/Xj5R4teeaxBzh+Jp6aVfzw8fbi8pVkugyeqVfDFCL2ylHff/89ixYtKlCdbt260b279aUTufnuu+84c+YMvr6+NG3alBMnTgBQpkyZXNcm+fn5ARAXF5ftc26/BHt4eODj40NiYiLx8fEEBATctM6NbeSXUwQpgLp16/Lcc8/x1Vdf8d133/F///d/ju6S2MDT05PPPp/Bip+Xs2zJDxw5cpiEhMtUCgjgkUfa8Wzv/8PXt6Td2qtUKYBFS8OZH/Ytq1et5OSJE7i5udKg4b089XRHQrt0w83Naf4aiJNLTbtO59dn0P2xJvTu2JR76wRQ2tebmNMX+XHtHj6bvz7bzuFm0xdsZMfeGF59tjXN7q9OverlSbqWxvodR5n703a+X5H76FDZkllTGAVZGP7LpgM81Os/DH2hPQ83rknDWpWIT7hK+Lq9TJr1S66vjxHnduHCBctIUUHqFMS2bdv48MMPAXj77bcpVqwYycnJQPZta27k4ZE1PW0ua950OT91zGXNdb28vKyWN1/LXC6/nOonyIABA/Dy8rI6tynOw2Aw8PgTT/H4E0/d0nX2Hjyar3JeXl680H8gL/QfeEvtidjL97/s5Ptfdt684D/8eegk/zd8ToHbGjplKUOnLC1wvYPRsTw/7JsC1xMHsNOQVNmyZalfv36B6+TX+vXrGTx4MOnp6fTo0YOuXbsCWEahrE3R3cj8wJl5oXh+6pjlt05BHmoDJwtSPj4+vPLKK47uhoiISKFhr4Xi3bt3L/A0XX6FhYXx/vvvYzQa6dWrF6NGjbKcK1asGEC2J/RvZH5bRdGiRQtcxzwCZa6TkpJitbz5WrmNWOXGqYKUiIiIOI+MjAzGjRvHwoULMRgMvPHGGwwcmH12wLxmKT4+HpPJZHXEyLxuybyOyVwnt6nFtLQ0rly5gsFgsIya3azOjW3kl4KUiIiIEyuMT+1B1gjPoEGD2Lx5M15eXkyePDnbPlBmvr6++Pv7c/78eU6cOEHVqlWznTcajRw7lrXHYO3atbN9joqKstp2VFQUJpOJypUrW0axblYnOjo6W7n8KvT7SImIiEjuCuMrYoxGoyVElS5dmrlz51oNUWbmDTBXr16d49yWLVtISkqiTp06lCtXDoAqVapQtWpV4uLi2LNnT446K1dmvZvyn9slNW/eHBcXFzZv3mx1QfmqVasAaNWqVb7vExSkRERExM6++OILNm/eTNGiRfn2229p2DDnJrT/1KtXL1xdXZk5c2a2YBQbG8v48eOBrAfO/un5558HYNSoUcTH/2+PtG3btjF37lzc3d3p27ev5bi/vz9t27YlISGBMWPGZHtdmbndGjVq0KZNmwLdq8FU0OXpd7nU3F9TJXLXKvmAHgIRsSZl97Tb3sauE1fscp37q/jcvFA+JCYm0rJlS5KTk6latSoNGjTIteyTTz5pGY2aNWsWU6ZMwdXVlaCgIDw9PdmxYwfJycl07do1x7t3MzMzeemll9iwYQPe3t4EBweTlJTEzp07MZlMTJ48mY4dO2arExcXR8+ePTl16hQVKlSgQYMGxMTEEBkZiY+PD/PmzSvw1J6CVAEpSInkpCAlYt2dCFK7TyTZ5Tr3VSlul+usWbMm30/YDxkyhH79+ln+e+3atcyZM4eDBw9aXiXTo0cPOnXqZHWzzuvXrzNv3jyWLl3KiRMn8Pb2pm7dugwYMICgoCCrbV66dInp06ezbt064uLi8PPzIygoiEGDBhEQEFDg+1WQKiAFKZGcFKRErLsbg9TdRk/tiYiIOLHC+tTe3UJBSkRExIkpRzmWgpSIiIgzU5JyKG1/ICIiImIjjUiJiIg4MXu9a09soyAlIiLixLTY3LE0tSciIiJiI41IiYiIODENSDmWgpSIiIgzU5JyKE3tiYiIiNhII1IiIiJOTE/tOZaClIiIiBPTU3uOpak9ERERERtpREpERMSJaUDKsRSkREREnJmSlEMpSImIiDgxLTZ3LK2REhEREbGRRqREREScmJ7acywFKRERESemHOVYmtoTERERsZFGpERERJyZhqQcSkFKRETEiempPcfS1J6IiIiIjTQiJSIi4sT01J5jKUiJiIg4MeUox9LUnoiIiIiNNCIlIiLizDQk5VAKUiIiIk5MT+05loKUiIiIE9Nic8fSGikRERERG2lESkRExIlpQMqxFKREREScmZKUQ2lqT0RERMRGGpESERFxYnpqz7EUpERERJyYntpzLE3tiYiIiNhII1IiIiJOTANSjqUgJSIi4sQ0tedYmtoTERERsZFGpERERJyahqQcSUFKRETEiWlqz7EUpERERJyYcpRjaY2UiIiIiI00IiUiIuLENLXnWApSIiIiTkyviHEsTe2JiIjIbXf8+HEaNWrEuHHjci2zdetW+vTpQ9OmTbnvvvvo3LkzixcvxmQyWS2fkZHBwoULCQ0N5f777ycoKIh+/fqxY8eOXNtITExkypQpPProozRs2JBmzZoxbNgwTp06ZdN9KUiJiIg4M4OdPm6j+Ph4Xn75ZVJSUnIts2DBAvr06cMff/xBvXr1CA4O5q+//mLkyJEMHz48R3mTycTQoUMZPXo0p0+fJiQkhFq1arF161Z69+7NkiVLctRJSEigZ8+ezJo1C6PRSMuWLSlZsiTLli2jY8eOHDlypMD3pqk9ERERJ1bYJ/YOHz7M66+/zokTJ3ItExMTw/jx4/H29iYsLIx69eoBEBsbS+/evVm6dCktWrTg0UcftdRZsmQJERER1K1blzlz5uDr6wvAtm3bGDBgAGPHjiUkJITy5ctb6kyaNIno6GhCQ0OZMGECrq6uAMyYMYOpU6fyzjvvEB4ejotL/seZNCIlIiIidpeYmMh//vMfunXrxokTJ6hUqVKuZWfPno3RaKRfv36WEAVQoUIFRo8ebSnzTzNnzgRg5MiRlhAF0LRpU3r37k1aWhphYWGW46dPnyYiIgIfHx9GjBhhCVEAAwcOpFGjRkRGRrJp06YC3aeClIiIiBMzGOzzYW9z585l9uzZlCpVii+++IKOHTvmWnb9+vUAtGvXLse5kJAQihcvzv79+zl//jwA0dHRnDx5ktKlS9O4ceMcddq3bw/A2rVrLcc2btyI0WgkJCQEb2/vfNXJDwUpERERJ2aw0//srVy5cgwdOpRVq1bRunXrXMvFx8dz8eJF3N3dCQwMzHHe1dXVcvzo0aMAREZGAlCzZk0MVlKg+fjJkyct67LMdWrVqmW1HzVq1AAo8DoprZESERERvv/+exYtWlSgOt26daN79+5Wz3Xt2jVf14iLiwOgTJkyua5N8vPzy1bW/Nnf399qeQ8PD3x8fEhMTCQ+Pp6AgICb1rmxjfxSkBIREXFmdhpMunDhAgcPHixwnVuVnJwMgKenZ65lPDw8spW9du1avuuYy5rrenl5WS1vvpa5XH4pSImIiDgxe03KlS1blvr16xe4zq0yj0JZm6K7kXk/KfNC8fzUMctvndz2rMqNgpSIiIgTs9dC8e7du+c6TXc7FStWDIDU1NRcy6SlpQFQtGjRAtcxj0CZ6+S2l5X5WrmNWOVGi81FRETEYcxrluLj43MdDTKvWzKvYzLXyW1qMS0tjStXrmAwGCyjZjerc2Mb+aUgJSIi4sQK61N7+eXr64u/vz/p6elWN+00Go0cO3YMgNq1a2f7HBUVZfWaUVFRmEwmAgICLKNYN6sTHR2drVx+KUiJiIg4scK6j1RBtGjRAoDVq1fnOLdlyxaSkpKoU6cO5cqVA6BKlSpUrVqVuLg49uzZk6POypUrAWjZsqXlWPPmzXFxcWHz5s1WF5SvWrUKgFatWhWo7wpSIiIi4lC9evXC1dWVmTNnZgtGsbGxjB8/HoABAwZkq/P8888DMGrUKOLj4y3Ht23bxty5c3F3d6dv376W4/7+/rRt25aEhATGjBlDenq65Zy53Ro1atCmTZsC9V2LzUVERMSh6tSpw+DBg5kyZQo9e/YkKCgIT09PduzYQXJyMl27dqVDhw7Z6vTo0YPffvuNDRs20L59e4KDg0lKSmLnzp2YTCYmT56c7T17kPU6mUOHDvHTTz+xc+dOGjRoQExMDJGRkfj4+PDxxx8X6ElAUJASERFxao6elrOX/v37U716debMmcO+ffswGAxUr16dHj160KlTpxzlXVxcmDZtGvPmzWPp0qVs3rwZb29vQkJCGDBgAEFBQTnq+Pn5sWjRIqZPn866detYt24dfn5+dOrUiUGDBhEQEFDgfhtMBd0w4S6XmuHoHogUPiUfeMXRXRAplFJ2T7vtbSSkGO1yHV8v15sXkhw0IiUiIuLEHPnEnShIiYiIOLV/y9Ses9JTeyIiIiI20oiUiIiIE9OAlGMpSImIiDgzJSmHUpASERFxYlps7lhaIyUiIiJiI41IiYiIODE9tedYClIiIiJOTDnKsTS1JyIiImIjjUiJiIg4Mw1JOZSClIiIiBPTU3uOpak9ERERERtpREpERMSJ6ak9xzKYTCaTozshIiIi4ow0tSciIiJiIwUpERERERspSImIiIjYSEFKRERExEYKUiIiIiI2UpASERERsZGClIiIiIiNFKREREREbKQgJSIiImIjBSkRERERGylIiYiIiNhIQUpERETERgpSIiIiIjZSkBIRERGxkYKUiIiIiI0UpERERERspCAlIiIiYiMFKXEqMTExvP3227Rq1YqGDRvSrl07pk6dSnJysqO7JlJoHD9+nEaNGjFu3DhHd0XkX09BSpzGvn37CA0NZfny5ZQpU4aWLVuSnJzMjBkz6N69O1evXnV0F0UcLj4+npdffpmUlBRHd0XkrqAgJU4hIyODN998k+TkZCZOnMjixYv59NNP+fXXX2ndujVHjx7l448/dnQ3RRzq8OHD9OzZk7/++svRXRG5ayhIiVP4+eefOXXqFE2bNqVLly6W456enkyaNImiRYuyaNEiEhMTHdhLEcdITEzkP//5D926dePEiRNUqlTJ0V0SuWsoSIlTWLduHQBt27bNca5kyZIEBwdz/fp1Nm3adKe7JuJwc+fOZfbs2ZQqVYovvviCjh07OrpLIncNBSlxCpGRkQDUrl3b6vkaNWoAcOTIkTvWJ5HColy5cgwdOpRVq1bRunVrR3dH5K7i5ugOiORHXFwcAP7+/lbP+/n5ZSsncjfp2rWro7sgctfSiJQ4BfP2Bp6enlbPm49rGwQREbmTFKTEKbi6ugJgMBjyLGcyme5Ed0RERAAFKXESxYoVA8h1b5zU1FQAvLy87lifREREFKTEKZjXQF24cMHqefPaKHM5ERGRO0FBSpyC+Wm96Ohoq+fNx3N7qk9EROR2UJASp9CiRQsAVq1alePc5cuX2bFjB+7u7jz00EN3umsiInIXU5ASp9C2bVsqVKjA5s2bmT9/vuV4amoqI0aMIDk5mS5dulCmTBkH9lJERO422kdKnIKnpyeTJ0+mf//+jBs3jiVLllCpUiV2795NXFwc9erV4+2333Z0N0VE5C6jESlxGsHBwSxevJj27dsTGxvLhg0bKF68OC+//DJhYWF4e3s7uosiInKXMZi08Y6IiIiITTQiJSIiImIjBSkRERERGylIiYiIiNhIQUpERETERgpSIiIiIjZSkBIRERGxkYKUiIiIiI0UpERERERspCAlIiIiYiMFKRFxWs899xy1a9dm6tSpju6KiNylFKREREREbKQgJSIiImIjBSkRERERGylIiYiIiNjIzdEdEJFb8/333zNmzBh8fX3ZvHkz7u7uOcqcPXuW1q1bYzAYWL9+Pf7+/ja11bp1a86cOUN4eDg//PAD4eHhZGRkUK1aNb7++mt8fX0BOHXqFF999RVbtmzh3LlzeHh4UKtWLTp16kRoaCiurq45rp2RkUFERAQrV67k4MGDJCQk4Obmhp+fH8HBwfTp04dq1arZ1G8RkdtFI1IiTu6JJ57A09OThIQENm7caLVMeHg4mZmZPPTQQzaHqH8aO3YsYWFh+Pn5UbJkSYoUKWIJUWvWrOGJJ55gwYIFxMXFERgYSKlSpfjzzz8ZOXIk/fr149q1a9mul5qaSt++fRk6dCjr16/H3d2dWrVqUaJECY4fP87ChQsJDQ3l0KFDt9x3ERF7UpAScXLe3t60a9cOgJ9++slqmR9//BGA0NBQu7S5a9cupk6dys8//8y6deuYPn06AEeOHOHNN98kLS2Nl156id9//53w8HBWr17NsmXLqFq1Ktu2beO9997Ldr0vv/ySHTt2ULJkSRYvXsy6detYsmQJGzZsYPHixZQtW5bk5GRmzJhhl/6LiNiLgpTIv0Dnzp0BWL9+PYmJidnO7dmzh5iYGHx9fWnTpo1d2mvSpAkdOnSw/HepUqUA+Oyzz0hPT+fZZ59l8ODBeHh4WMrUq1ePTz/9FFdXV5YvX050dLTl3NatW3FxceGVV16hYcOG2dpq2LAhPXr0ACAyMtIu/RcRsRcFKZF/geDgYAICAkhPT+eXX37Jdm7ZsmVA1hRgkSJF7NJe48aNcxxLT0/nt99+A+Cpp56yWq927drUqVMHk8nE+vXrLccXLFjAvn376N69u9V6Xl5eQNYUoIhIYaLF5iL/AgaDgU6dOvHpp58SHh5uCST/DFb2mtYDKFu2bI5jx48fJz09HchaQ5VbaIuNjQXg2LFj2Y67u7uTlJTErl27OH78OKdOneL48eMcPnyY+Ph4ADIzM+12DyIi9qAgJfIvERoayrRp09i1axenTp0iICCAdevWkZiYSO3atalfv77d2vL09MxxLCkpyfL1gQMHbnqNf5a/evUqH3/8McuWLSM5Odly3N3dnfr161O3bl02bdp0i70WEbE/BSmRf4ny5csTEhLC5s2bWb58OS+//DLh4eGAfUejclO0aFHL17t27aJYsWL5rvvyyy+zY8cOPD096dOnD/feey81a9akSpUquLu7s2jRIgUpESmUtEZK5F/EvOh8zZo1pKSksGXLFtzd3XNds2RPAQEBlv2h/rmQ/Eb79u3j6NGjli0Q9uzZw44dOwCYOXMmw4YN47HHHqNGjRqWPbHOnTt3m3svImIbBSmRf5FHHnkEX19fDh06xMKFC0lLS6Nly5aWp+puJ29vb4KCggCYO3eu1TKnTp2iZ8+ePPXUU6xcuRKA06dPW87fc889OeqkpKTw888/A2A0Gu3dbRGRW6IgJfIvUqRIEZ588kkA/vvf/wJ3ZlrP7NVXX8XV1ZWIiAjef//9bBtvRkZG0r9/f65fv07FihUt/QwMDLSU+fzzz8nIyLD8d3R0NC+++CLHjx8HskKViEhhoiAl8i/TpUsXAJKTkylTpgzNmze/Y203btyY8ePH4+7uzpw5cwgJCSE0NJT27dvz1FNPcezYMcqUKcNXX31leaqvXr16PPbYYwB8/fXXNGvWjM6dO9OmTRsef/xx/vjjDx566CEArl27xtWrV+/Y/YiI3IyClMi/TJ06dahTpw6QtZ+Tm9udfaakc+fOhIeH88wzz1C2bFmioqI4e/Ys1atX54UXXuCnn37K8c68KVOmMH78eBo0aEBmZiZHjx4lPT2dVq1aMXPmTL7++msqVqwIwLp16+7o/YiI5MVgMplMju6EiNhPRkYGLVq0ID4+np9//pkaNWo4uksiIv9aGpES+ZdZt24d8fHx3HfffQpRIiK3mfaREvkXOHbsGB4eHsTExDB27FgA+vTp4+BeiYj8+ylIifwLzJkzh4ULF1r+u1mzZrRv3z5Huddee40LFy4U+Pr16tVj1KhRt9RHEZF/IwUpkX+BBg0asHr1aoxGI4888ggjRoywWu7AgQOcOXOmwNe/0wvWRUSchRabi4iIiNhIi81FREREbKQgJSIiImIjBSkRERERGylIiYiIiNhIQUpERETERgpSIiIiIjZSkBIRERGxkYKUiIiIiI3+H7ouoOaGgqwDAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ds_validacion=pd.DataFrame(y_predCerteza,y_test).reset_index()\n",
    "ds_validacion.columns=['y_pred','y_real']\n",
    "\n",
    "tabla=pd.crosstab(ds_validacion.y_pred, ds_validacion.y_real)\n",
    "grf=sns.heatmap(tabla,annot=True, cmap = 'Blues', fmt='g')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "ba944be8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.8176170095264393\n",
      "Recall: 0.9142777623156136\n",
      "Precision: 0.7643089809213588\n",
      "f1 score: 0.8325940945380813\n"
     ]
    }
   ],
   "source": [
    "#Calculo las métricas en el conjunto de evaluación\n",
    "accuracy=accuracy_score(y_test,y_predCerteza)\n",
    "recall=recall_score(y_test,y_predCerteza)\n",
    "f1=f1_score(y_test,y_predCerteza,)\n",
    "precision=precision_score(y_test,y_predCerteza)\n",
    "\n",
    "print(\"Accuracy: \"+str(accuracy))\n",
    "print(\"Recall: \"+str(recall))\n",
    "print(\"Precision: \"+str(precision))\n",
    "print(\"f1 score: \"+str(f1))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bf453ebd",
   "metadata": {},
   "source": [
    "### Prediccion al set de datos de test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "b7faede7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0],\n",
       "       [1],\n",
       "       [0],\n",
       "       ...,\n",
       "       [0],\n",
       "       [0],\n",
       "       [0]])"
      ]
     },
     "execution_count": 70,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "if exists('modelos/TP2/predTestRedConOptimi.joblib') == False:\n",
    "    y_pred = modeloHiper.predict(data_list_submit)\n",
    "    y_predCerteza = np.where(y_pred>0.7,1,0)\n",
    "    dump(y_predCerteza, 'modelos/TP2/predTestRedConOptimi.joblib')\n",
    "else:\n",
    "    y_predCerteza = load('modelos/TP2/predTestRedConOptimi.joblib')\n",
    "y_predCerteza"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "4e6f6e41",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('submissions/TP2/redesNeuronales3.csv'):\n",
    "    yEnEspanol =y_predCerteza\n",
    "    yEnEspanol = np.array(list(map(lambda x: \"positivo\" if x==1 else \"negativo\", y_predCerteza)))\n",
    "    df_submission = pd.DataFrame({'id': review_pruebasOriginal['ID'], 'sentimiento': yEnEspanol})\n",
    "    df_submission.to_csv('submissions/TP2/redesNeuronales3.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b3c39dcb",
   "metadata": {},
   "source": [
    "# Ensamble de 3 modelos\n",
    "\n",
    "Armamos un ensamble sencillo de tipo voting con los modelos bases generados a lo largo del analisis.\n",
    "\n",
    "La utilización de los modelos bases se debe a que utilizamos varios predictores debiles con el proposito de construir uno que en conjunto funcione mejor y asi aprovechar tambien los modelos diseñados previamente "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "e1c4a3f2",
   "metadata": {
    "lines_to_next_cell": 0
   },
   "outputs": [],
   "source": [
    "if not exists('modelos/TP2/voting.joblib.gz'): #Tenemos el archivo comprimido porque era muy grande\n",
    "    naives = load(\"modelos/TP2/nb_conhp.joblib\")\n",
    "    xgb = load(\"modelos/TP2/xgb_base.joblib\")\n",
    "    rf = load(\"modelos/TP2/modeloRandomForest-sin-optimizar-final.joblib\")\n",
    "\n",
    "    ensamble = VotingClassifier(estimators = [\n",
    "        ('nb', naives),\n",
    "        ('xgb', xgb),\n",
    "        ('rf', rf)\n",
    "    ])\n",
    "    ensamble.fit(x_train, y_train)\n",
    "    dump(ensamble, 'modelos/TP2/voting.joblib')\n",
    "    !gzip modelos/TP2/voting.joblib #Comprimimos\n",
    "\n",
    "else:\n",
    "    !gzip -d -k modelos/TP2/voting.joblib.gz #Descomprimimos\n",
    "    ensamble = load('modelos/TP2/voting.joblib')\n",
    "    !rm modelos/TP2/voting.joblib\n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e30e092e",
   "metadata": {},
   "source": [
    "Entrenamos el modelo "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "7ec6727e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-2 {color: black;background-color: white;}#sk-container-id-2 pre{padding: 0;}#sk-container-id-2 div.sk-toggleable {background-color: white;}#sk-container-id-2 label.sk-toggleable__label {cursor: pointer;display: block;width: 100%;margin-bottom: 0;padding: 0.3em;box-sizing: border-box;text-align: center;}#sk-container-id-2 label.sk-toggleable__label-arrow:before {content: \"▸\";float: left;margin-right: 0.25em;color: #696969;}#sk-container-id-2 label.sk-toggleable__label-arrow:hover:before {color: black;}#sk-container-id-2 div.sk-estimator:hover label.sk-toggleable__label-arrow:before {color: black;}#sk-container-id-2 div.sk-toggleable__content {max-height: 0;max-width: 0;overflow: hidden;text-align: left;background-color: #f0f8ff;}#sk-container-id-2 div.sk-toggleable__content pre {margin: 0.2em;color: black;border-radius: 0.25em;background-color: #f0f8ff;}#sk-container-id-2 input.sk-toggleable__control:checked~div.sk-toggleable__content {max-height: 200px;max-width: 100%;overflow: auto;}#sk-container-id-2 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {content: \"▾\";}#sk-container-id-2 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-2 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-2 input.sk-hidden--visually {border: 0;clip: rect(1px 1px 1px 1px);clip: rect(1px, 1px, 1px, 1px);height: 1px;margin: -1px;overflow: hidden;padding: 0;position: absolute;width: 1px;}#sk-container-id-2 div.sk-estimator {font-family: monospace;background-color: #f0f8ff;border: 1px dotted black;border-radius: 0.25em;box-sizing: border-box;margin-bottom: 0.5em;}#sk-container-id-2 div.sk-estimator:hover {background-color: #d4ebff;}#sk-container-id-2 div.sk-parallel-item::after {content: \"\";width: 100%;border-bottom: 1px solid gray;flex-grow: 1;}#sk-container-id-2 div.sk-label:hover label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-2 div.sk-serial::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: 0;}#sk-container-id-2 div.sk-serial {display: flex;flex-direction: column;align-items: center;background-color: white;padding-right: 0.2em;padding-left: 0.2em;position: relative;}#sk-container-id-2 div.sk-item {position: relative;z-index: 1;}#sk-container-id-2 div.sk-parallel {display: flex;align-items: stretch;justify-content: center;background-color: white;position: relative;}#sk-container-id-2 div.sk-item::before, #sk-container-id-2 div.sk-parallel-item::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: -1;}#sk-container-id-2 div.sk-parallel-item {display: flex;flex-direction: column;z-index: 1;position: relative;background-color: white;}#sk-container-id-2 div.sk-parallel-item:first-child::after {align-self: flex-end;width: 50%;}#sk-container-id-2 div.sk-parallel-item:last-child::after {align-self: flex-start;width: 50%;}#sk-container-id-2 div.sk-parallel-item:only-child::after {width: 0;}#sk-container-id-2 div.sk-dashed-wrapped {border: 1px dashed gray;margin: 0 0.4em 0.5em 0.4em;box-sizing: border-box;padding-bottom: 0.4em;background-color: white;}#sk-container-id-2 div.sk-label label {font-family: monospace;font-weight: bold;display: inline-block;line-height: 1.2em;}#sk-container-id-2 div.sk-label-container {text-align: center;}#sk-container-id-2 div.sk-container {/* jupyter's `normalize.less` sets `[hidden] { display: none; }` but bootstrap.min.css set `[hidden] { display: none !important; }` so we also need the `!important` here to be able to override the default hidden behavior on the sphinx rendered scikit-learn.org. See: https://github.com/scikit-learn/scikit-learn/issues/21755 */display: inline-block !important;position: relative;}#sk-container-id-2 div.sk-text-repr-fallback {display: none;}</style><div id=\"sk-container-id-2\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>VotingClassifier(estimators=[(&#x27;nb&#x27;,\n",
       "                              Pipeline(steps=[(&#x27;tfidfvectorizer&#x27;,\n",
       "                                               TfidfVectorizer(max_df=0.9,\n",
       "                                                               min_df=0.03,\n",
       "                                                               ngram_range=(1,\n",
       "                                                                            2),\n",
       "                                                               stop_words=&#x27;english&#x27;)),\n",
       "                                              (&#x27;multinomialnb&#x27;,\n",
       "                                               MultinomialNB(alpha=0.3))])),\n",
       "                             (&#x27;xgb&#x27;,\n",
       "                              Pipeline(steps=[(&#x27;tfidfvectorizer&#x27;,\n",
       "                                               TfidfVectorizer()),\n",
       "                                              (&#x27;xgbclassifier&#x27;,\n",
       "                                               XGBClassifier(base_score=0.5,\n",
       "                                                             booster=&#x27;gbtree&#x27;,\n",
       "                                                             callbacks=None,\n",
       "                                                             colsample_bylev...\n",
       "                                                             monotone_constraints=&#x27;()&#x27;,\n",
       "                                                             n_estimators=100,\n",
       "                                                             n_jobs=0,\n",
       "                                                             num_parallel_tree=1,\n",
       "                                                             predictor=&#x27;auto&#x27;,\n",
       "                                                             random_state=0,\n",
       "                                                             reg_alpha=0,\n",
       "                                                             reg_lambda=1, ...))])),\n",
       "                             (&#x27;rf&#x27;,\n",
       "                              Pipeline(steps=[(&#x27;tfidfvectorizer&#x27;,\n",
       "                                               TfidfVectorizer()),\n",
       "                                              (&#x27;randomforestclassifier&#x27;,\n",
       "                                               RandomForestClassifier(class_weight=&#x27;balanced&#x27;,\n",
       "                                                                      min_samples_leaf=15,\n",
       "                                                                      min_samples_split=40,\n",
       "                                                                      n_estimators=40,\n",
       "                                                                      n_jobs=-2,\n",
       "                                                                      random_state=9))]))])</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-5\" type=\"checkbox\" ><label for=\"sk-estimator-id-5\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">VotingClassifier</label><div class=\"sk-toggleable__content\"><pre>VotingClassifier(estimators=[(&#x27;nb&#x27;,\n",
       "                              Pipeline(steps=[(&#x27;tfidfvectorizer&#x27;,\n",
       "                                               TfidfVectorizer(max_df=0.9,\n",
       "                                                               min_df=0.03,\n",
       "                                                               ngram_range=(1,\n",
       "                                                                            2),\n",
       "                                                               stop_words=&#x27;english&#x27;)),\n",
       "                                              (&#x27;multinomialnb&#x27;,\n",
       "                                               MultinomialNB(alpha=0.3))])),\n",
       "                             (&#x27;xgb&#x27;,\n",
       "                              Pipeline(steps=[(&#x27;tfidfvectorizer&#x27;,\n",
       "                                               TfidfVectorizer()),\n",
       "                                              (&#x27;xgbclassifier&#x27;,\n",
       "                                               XGBClassifier(base_score=0.5,\n",
       "                                                             booster=&#x27;gbtree&#x27;,\n",
       "                                                             callbacks=None,\n",
       "                                                             colsample_bylev...\n",
       "                                                             monotone_constraints=&#x27;()&#x27;,\n",
       "                                                             n_estimators=100,\n",
       "                                                             n_jobs=0,\n",
       "                                                             num_parallel_tree=1,\n",
       "                                                             predictor=&#x27;auto&#x27;,\n",
       "                                                             random_state=0,\n",
       "                                                             reg_alpha=0,\n",
       "                                                             reg_lambda=1, ...))])),\n",
       "                             (&#x27;rf&#x27;,\n",
       "                              Pipeline(steps=[(&#x27;tfidfvectorizer&#x27;,\n",
       "                                               TfidfVectorizer()),\n",
       "                                              (&#x27;randomforestclassifier&#x27;,\n",
       "                                               RandomForestClassifier(class_weight=&#x27;balanced&#x27;,\n",
       "                                                                      min_samples_leaf=15,\n",
       "                                                                      min_samples_split=40,\n",
       "                                                                      n_estimators=40,\n",
       "                                                                      n_jobs=-2,\n",
       "                                                                      random_state=9))]))])</pre></div></div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><label>nb</label></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-6\" type=\"checkbox\" ><label for=\"sk-estimator-id-6\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">TfidfVectorizer</label><div class=\"sk-toggleable__content\"><pre>TfidfVectorizer(max_df=0.9, min_df=0.03, ngram_range=(1, 2),\n",
       "                stop_words=&#x27;english&#x27;)</pre></div></div></div><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-7\" type=\"checkbox\" ><label for=\"sk-estimator-id-7\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">MultinomialNB</label><div class=\"sk-toggleable__content\"><pre>MultinomialNB(alpha=0.3)</pre></div></div></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><label>xgb</label></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-8\" type=\"checkbox\" ><label for=\"sk-estimator-id-8\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">TfidfVectorizer</label><div class=\"sk-toggleable__content\"><pre>TfidfVectorizer()</pre></div></div></div><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-9\" type=\"checkbox\" ><label for=\"sk-estimator-id-9\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">XGBClassifier</label><div class=\"sk-toggleable__content\"><pre>XGBClassifier(base_score=0.5, booster=&#x27;gbtree&#x27;, callbacks=None,\n",
       "              colsample_bylevel=1, colsample_bynode=1, colsample_bytree=1,\n",
       "              early_stopping_rounds=None, enable_categorical=False,\n",
       "              eval_metric=None, gamma=0, gpu_id=-1, grow_policy=&#x27;depthwise&#x27;,\n",
       "              importance_type=None, interaction_constraints=&#x27;&#x27;,\n",
       "              learning_rate=0.300000012, max_bin=256, max_cat_to_onehot=4,\n",
       "              max_delta_step=0, max_depth=6, max_leaves=0, min_child_weight=1,\n",
       "              missing=nan, monotone_constraints=&#x27;()&#x27;, n_estimators=100,\n",
       "              n_jobs=0, num_parallel_tree=1, predictor=&#x27;auto&#x27;, random_state=0,\n",
       "              reg_alpha=0, reg_lambda=1, ...)</pre></div></div></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><label>rf</label></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-10\" type=\"checkbox\" ><label for=\"sk-estimator-id-10\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">TfidfVectorizer</label><div class=\"sk-toggleable__content\"><pre>TfidfVectorizer()</pre></div></div></div><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-11\" type=\"checkbox\" ><label for=\"sk-estimator-id-11\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">RandomForestClassifier</label><div class=\"sk-toggleable__content\"><pre>RandomForestClassifier(class_weight=&#x27;balanced&#x27;, min_samples_leaf=15,\n",
       "                       min_samples_split=40, n_estimators=40, n_jobs=-2,\n",
       "                       random_state=9)</pre></div></div></div></div></div></div></div></div></div></div></div></div>"
      ],
      "text/plain": [
       "VotingClassifier(estimators=[('nb',\n",
       "                              Pipeline(steps=[('tfidfvectorizer',\n",
       "                                               TfidfVectorizer(max_df=0.9,\n",
       "                                                               min_df=0.03,\n",
       "                                                               ngram_range=(1,\n",
       "                                                                            2),\n",
       "                                                               stop_words='english')),\n",
       "                                              ('multinomialnb',\n",
       "                                               MultinomialNB(alpha=0.3))])),\n",
       "                             ('xgb',\n",
       "                              Pipeline(steps=[('tfidfvectorizer',\n",
       "                                               TfidfVectorizer()),\n",
       "                                              ('xgbclassifier',\n",
       "                                               XGBClassifier(base_score=0.5,\n",
       "                                                             booster='gbtree',\n",
       "                                                             callbacks=None,\n",
       "                                                             colsample_bylev...\n",
       "                                                             monotone_constraints='()',\n",
       "                                                             n_estimators=100,\n",
       "                                                             n_jobs=0,\n",
       "                                                             num_parallel_tree=1,\n",
       "                                                             predictor='auto',\n",
       "                                                             random_state=0,\n",
       "                                                             reg_alpha=0,\n",
       "                                                             reg_lambda=1, ...))])),\n",
       "                             ('rf',\n",
       "                              Pipeline(steps=[('tfidfvectorizer',\n",
       "                                               TfidfVectorizer()),\n",
       "                                              ('randomforestclassifier',\n",
       "                                               RandomForestClassifier(class_weight='balanced',\n",
       "                                                                      min_samples_leaf=15,\n",
       "                                                                      min_samples_split=40,\n",
       "                                                                      n_estimators=40,\n",
       "                                                                      n_jobs=-2,\n",
       "                                                                      random_state=9))]))])"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ensamble"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e3343af5",
   "metadata": {},
   "source": [
    "Sobre el ensamble realizamos una validacion cruzada para comprobar que su metricas al estar asociadas a las metricas individuales de los modelos siguie siendo optima "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "e2874941",
   "metadata": {},
   "outputs": [],
   "source": [
    "kfoldcv = StratifiedKFold(n_splits=5) \n",
    "scorer_fn = make_scorer(sklearn.metrics.f1_score)\n",
    "\n",
    "if not exists ('modelos/TP2/ensambleCrossvalidation.joblib.gz.part-aa'):\n",
    "    resultados_rf = cross_validate(ensamble, x_train, \n",
    "                                   y_train, cv=kfoldcv,\n",
    "                                   scoring=scorer_fn,\n",
    "                                   return_estimator=True, n_jobs = JOBS)\n",
    "    \n",
    "    dump(resultados_rf, \"modelos/TP2/ensambleCrossvalidation.joblib\")\n",
    "    #Comprimimos el archivo\n",
    "    !gzip -v -9 modelos/TP2/ensambleCrossvalidation.joblib\n",
    "    #Separamos en varios archivos para que github no nos haga drama. #GithubDestruidoConElPoderDeUnix\n",
    "    !split -b 50m modelos/TP2/ensambleCrossvalidation.joblib.gz modelos/TP2/ensambleCrossvalidation.joblib.gz.part-\n",
    "    !rm modelos/TP2/ensambleCrossvalidation.joblib.gz\n",
    "\n",
    "else:\n",
    "    #Reconstruimos el zip\n",
    "    !cat modelos/TP2/ensambleCrossvalidation.joblib.gz.part-a* > modelos/TP2/ensambleCrossvalidation.joblib.gz\n",
    "    #Descomprimimos\n",
    "    !gzip -d -k modelos/TP2/ensambleCrossvalidation.joblib.gz\n",
    "    resultados_rf = load('modelos/TP2/ensambleCrossvalidation.joblib')\n",
    "    !rm modelos/TP2/ensambleCrossvalidation.joblib\n",
    "    !rm modelos/TP2/ensambleCrossvalidation.joblib.gz\n",
    "\n",
    "\n",
    "metricas_cv_rf = resultados_rf['test_score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "101a03d6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.83010753, 0.83674358, 0.83285303, 0.83278026, 0.83367556])"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metricas_cv_rf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "9287b987",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Modelo entrenado con 5 folds')"
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "metric_labels_CV_rf = ['F1 Score']*len(metricas_cv_rf) \n",
    "sns.set_context('talk')\n",
    "sns.set_style(\"darkgrid\")\n",
    "plt.figure(figsize=(8,8))\n",
    "sns.boxplot(metricas_cv_rf)\n",
    "plt.title(\"Modelo entrenado con 5 folds\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "82e5fb37",
   "metadata": {},
   "source": [
    "## Hacemos prediccion sobre el set de testeo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "2ee6eb03",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not exists('submissions/TP2/ensamble.csv'):\n",
    "    pred = ensamble.predict(review_pruebas.review_es)\n",
    "    df_submission = pd.DataFrame({'id': review_pruebasOriginal['ID'], 'sentimiento': pred})\n",
    "    df_submission['sentimiento'] = df_submission['sentimiento'].map({1: 'positivo', 0: 'negativo'})\n",
    "    df_submission.to_csv('submissions/TP2/ensamble.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7a737470",
   "metadata": {},
   "source": [
    "Nota sobre el ensamble hibrido: \n",
    "\n",
    "El modelo de mejor performance en kaggle que elevo la posicion del equipo es un modelo incompleto puesto que esta armado unicamente con dos modelos, pero, el modelo expuesto en esta porción del analisis incluye los dos modelos usados en el armado del ensamble que genero la mejor predicción\n",
    "\n",
    "De manera conclusiva es importante destacar que los ensambles son los modelos que mejor encaran la predicción en el analisis de sentimientos de las reviews de peliculas "
   ]
  }
 ],
 "metadata": {
  "jupytext": {
   "cell_metadata_filter": "-all",
   "formats": "ipynb,md"
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}