Notebook_AnaliseSentimentosGovernoMinas_MinerandoDados.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Análise de Sentimentos – Aprenda de uma vez por todas como funciona utilizando dados do Twitter**\n",
    "\n",
    "                                    wwww.minerandodados.com.br"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import nltk\n",
    "import re\n",
    "import pandas as pd\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn import metrics\n",
    "from sklearn.model_selection import cross_val_predict"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Ler arquivo de dados e conta a quantidade de linhas**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dataset = pd.read_csv('Tweets_Mg.csv',encoding='utf-8')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Unnamed: 0                   8199\n",
       "Created At                   8199\n",
       "Text                         8199\n",
       "Geo Coordinates.latitude      104\n",
       "Geo Coordinates.longitude     104\n",
       "User Location                5489\n",
       "Username                     8199\n",
       "User Screen Name             8199\n",
       "Retweet Count                8199\n",
       "Classificacao                8199\n",
       "Observação                      1\n",
       "Unnamed: 10                     0\n",
       "Unnamed: 11                     0\n",
       "Unnamed: 12                     0\n",
       "Unnamed: 13                     0\n",
       "Unnamed: 14                     0\n",
       "Unnamed: 15                     0\n",
       "Unnamed: 16                     0\n",
       "Unnamed: 17                     0\n",
       "Unnamed: 18                     0\n",
       "Unnamed: 19                     0\n",
       "Unnamed: 20                     0\n",
       "Unnamed: 21                     0\n",
       "Unnamed: 22                     0\n",
       "Unnamed: 23                     0\n",
       "Unnamed: 24                     0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Conta a quantidade de linhas de tweets neutros, positivos e negativos**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 206,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Unnamed: 0                   2453\n",
       "Created At                   2453\n",
       "Text                         2453\n",
       "Geo Coordinates.latitude      102\n",
       "Geo Coordinates.longitude     102\n",
       "User Location                1712\n",
       "Username                     2453\n",
       "User Screen Name             2453\n",
       "Retweet Count                2453\n",
       "Classificacao                2453\n",
       "Observação                      0\n",
       "Unnamed: 10                     0\n",
       "Unnamed: 11                     0\n",
       "Unnamed: 12                     0\n",
       "Unnamed: 13                     0\n",
       "Unnamed: 14                     0\n",
       "Unnamed: 15                     0\n",
       "Unnamed: 16                     0\n",
       "Unnamed: 17                     0\n",
       "Unnamed: 18                     0\n",
       "Unnamed: 19                     0\n",
       "Unnamed: 20                     0\n",
       "Unnamed: 21                     0\n",
       "Unnamed: 22                     0\n",
       "Unnamed: 23                     0\n",
       "Unnamed: 24                     0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 206,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset[dataset.Classificacao=='Neutro'].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 207,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Unnamed: 0                   3300\n",
       "Created At                   3300\n",
       "Text                         3300\n",
       "Geo Coordinates.latitude        1\n",
       "Geo Coordinates.longitude       1\n",
       "User Location                2118\n",
       "Username                     3300\n",
       "User Screen Name             3300\n",
       "Retweet Count                3300\n",
       "Classificacao                3300\n",
       "Observação                      1\n",
       "Unnamed: 10                     0\n",
       "Unnamed: 11                     0\n",
       "Unnamed: 12                     0\n",
       "Unnamed: 13                     0\n",
       "Unnamed: 14                     0\n",
       "Unnamed: 15                     0\n",
       "Unnamed: 16                     0\n",
       "Unnamed: 17                     0\n",
       "Unnamed: 18                     0\n",
       "Unnamed: 19                     0\n",
       "Unnamed: 20                     0\n",
       "Unnamed: 21                     0\n",
       "Unnamed: 22                     0\n",
       "Unnamed: 23                     0\n",
       "Unnamed: 24                     0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 207,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset[dataset.Classificacao=='Positivo'].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Unnamed: 0                   2446\n",
       "Created At                   2446\n",
       "Text                         2446\n",
       "Geo Coordinates.latitude        1\n",
       "Geo Coordinates.longitude       1\n",
       "User Location                1659\n",
       "Username                     2446\n",
       "User Screen Name             2446\n",
       "Retweet Count                2446\n",
       "Classificacao                2446\n",
       "Observação                      0\n",
       "Unnamed: 10                     0\n",
       "Unnamed: 11                     0\n",
       "Unnamed: 12                     0\n",
       "Unnamed: 13                     0\n",
       "Unnamed: 14                     0\n",
       "Unnamed: 15                     0\n",
       "Unnamed: 16                     0\n",
       "Unnamed: 17                     0\n",
       "Unnamed: 18                     0\n",
       "Unnamed: 19                     0\n",
       "Unnamed: 20                     0\n",
       "Unnamed: 21                     0\n",
       "Unnamed: 22                     0\n",
       "Unnamed: 23                     0\n",
       "Unnamed: 24                     0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset[dataset.Classificacao=='Negativo'].count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Pre-Processamento - Lembre-se: Nesse artigo NÃO UTILIZAMOS essas funções, conforme explicado. **"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def PreprocessamentoSemStopWords(instancia):\n",
    "    #remove links dos tweets\n",
    "    #remove stopwords\n",
    "    instancia = re.sub(r\"http\\S+\", \"\", instancia).lower().replace(',','').replace('.','').replace(';','').replace('-','')\n",
    "    stopwords = set(nltk.corpus.stopwords.words('portuguese'))\n",
    "    palavras = [i for i in instancia.split() if not i in stopwords]\n",
    "    return (\" \".join(palavras))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def Stemming(instancia):\n",
    "    stemmer = nltk.stem.RSLPStemmer()\n",
    "    palavras=[]\n",
    "    for w in instancia.split():\n",
    "        palavras.append(stemmer.stem(w))\n",
    "    return (\" \".join(palavras))    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def Preprocessamento(instancia):\n",
    "    #remove links, pontos, virgulas,ponto e virgulas dos tweets\n",
    "    #coloca tudo em minusculo\n",
    "    instancia = re.sub(r\"http\\S+\", \"\", instancia).lower().replace(',','').replace('.','').replace(';','').replace('-','').replace(':','')\n",
    "    return (instancia)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'gosto partido votaria novamente nesse governante!'"
      ]
     },
     "execution_count": 191,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Preprocessamento_SemStopWords('Eu não gosto do partido, e também não votaria novamente nesse governante!')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Separando tweets e suas classes\n",
    "**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "tweets = dataset['Text'].values\n",
    "classes = dataset['Classificacao'].values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "**Gerando o modelo**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB(alpha=1.0, class_prior=None, fit_prior=True)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vectorizer = CountVectorizer(analyzer=\"word\")\n",
    "freq_tweets = vectorizer.fit_transform(tweets)\n",
    "modelo = MultinomialNB()\n",
    "modelo.fit(freq_tweets,classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Testando o modelo com algumas instâncias simples**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# defina instâncias de teste dentro de uma lista"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "testes = ['Esse governo está no início, vamos ver o que vai dar','Estou muito feliz com o governo de Minas esse ano','O estado de Minas Gerais decretou calamidade financeira!!!','A segurança desse país está deixando a desejar','O governador de Minas é do PT']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "freq_testes = vectorizer.transform(testes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Fazendo a classificação com o modelo treinado."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Neutro', 'Neutro', 'Negativo', 'Negativo', 'Neutro'], \n",
       "      dtype='<U8')"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "modelo.predict(freq_testes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Avaliando o modelo **"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 214,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Fazendo o cross validation do modelo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "resultados = cross_val_predict(modelo, freq_tweets, classes, cv=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Medindo a acurácia média do modelo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8831564824978656"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metrics.accuracy_score(classes,resultados)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Medidas de validação do modelo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "   Positivo       0.95      0.88      0.91      3300\n",
      "   Negativo       0.89      0.93      0.91      2446\n",
      "     Neutro       0.80      0.84      0.82      2453\n",
      "\n",
      "avg / total       0.89      0.88      0.88      8199\n",
      "\n"
     ]
    }
   ],
   "source": [
    "sentimento=['Positivo','Negativo','Neutro']\n",
    "print (metrics.classification_report(classes,resultados,sentimento))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Matriz de confusão"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predito   Negativo  Neutro  Positivo   All\n",
      "Real                                      \n",
      "Negativo      2275     162         9  2446\n",
      "Neutro         240    2067       146  2453\n",
      "Positivo        45     356      2899  3300\n",
      "All           2560    2585      3054  8199\n"
     ]
    }
   ],
   "source": [
    "print (pd.crosstab(classes, resultados, rownames=['Real'], colnames=['Predito'], margins=True))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Melhorando resultados com Bigrams**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB(alpha=1.0, class_prior=None, fit_prior=True)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vectorizer = CountVectorizer(ngram_range=(1,2))\n",
    "freq_tweets = vectorizer.fit_transform(tweets)\n",
    "modelo = MultinomialNB()\n",
    "modelo.fit(freq_tweets,classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "resultados = cross_val_predict(modelo, freq_tweets, classes, cv=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.89547505793389437"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metrics.accuracy_score(classes,resultados)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "   Positivo       0.97      0.88      0.92      3300\n",
      "   Negativo       0.91      0.93      0.92      2446\n",
      "     Neutro       0.80      0.89      0.84      2453\n",
      "\n",
      "avg / total       0.90      0.90      0.90      8199\n",
      "\n"
     ]
    }
   ],
   "source": [
    "sentimento=['Positivo','Negativo','Neutro']\n",
    "print (metrics.classification_report(classes,resultados,sentimento))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predito   Negativo  Neutro  Positivo   All\n",
      "Real                                      \n",
      "Negativo      2265     179         2  2446\n",
      "Neutro         181    2177        95  2453\n",
      "Positivo        43     357      2900  3300\n",
      "All           2489    2713      2997  8199\n"
     ]
    }
   ],
   "source": [
    "print (pd.crosstab(classes, resultados, rownames=['Real'], colnames=['Predito'], margins=True))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}