banknote_classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Introduction\n",
    "\n",
    "The notebook contains a solution to the binary classification problem, based on a dataset available under [this link](https://archive.ics.uci.edu/ml/machine-learning-databases/00267/). The idea of the task was to distinguish genuine and forged banknotes based on certain features of their images. \n",
    "\n",
    "The provided solution served as an example to test and present a few of the most important performance metrics for binary classifiers. Stochastic gradient descent classifier was used.\n",
    "\n",
    "The code was based on Chapter 3 of Aurelien Geron, 'Hands - On Machine Learing with Scikit-Learn and TensorFlow', Helion SA, 2018. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "from six.moves import urllib\n",
    "import tarfile\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "DOWNLOAD_ROOT = 'http://archive.ics.uci.edu/ml/machine-learning-databases/'\n",
    "BANKNOTE_PATH = os.path.join('datasets', 'banknotes')\n",
    "BANKNOTE_URL = DOWNLOAD_ROOT + '00267/data_banknote_authentication.txt'\n",
    "\n",
    "def fetch_data(data_url, data_path, filename, is_tar = False):\n",
    "    if  not os.path.isdir(data_path):\n",
    "        os.makedirs(data_path)\n",
    "    file_path = os.path.join(data_path, filename)\n",
    "    urllib.request.urlretrieve(data_url, file_path)\n",
    "    if is_tar:\n",
    "        with tarfile.TarFile(file_path) as current_tar:\n",
    "            current_tar.extractall(data_path)\n",
    "\n",
    "def load_data(data_path, filename, separator = \",\", header = None):\n",
    "    file_path = os.path.join(data_path, filename)\n",
    "    return pd.read_csv(file_path, sep = separator, header = header)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3.62160</td>\n",
       "      <td>8.6661</td>\n",
       "      <td>-2.8073</td>\n",
       "      <td>-0.44699</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4.54590</td>\n",
       "      <td>8.1674</td>\n",
       "      <td>-2.4586</td>\n",
       "      <td>-1.46210</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3.86600</td>\n",
       "      <td>-2.6383</td>\n",
       "      <td>1.9242</td>\n",
       "      <td>0.10645</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3.45660</td>\n",
       "      <td>9.5228</td>\n",
       "      <td>-4.0112</td>\n",
       "      <td>-3.59440</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.32924</td>\n",
       "      <td>-4.4552</td>\n",
       "      <td>4.5718</td>\n",
       "      <td>-0.98880</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         1       2       3        4  label\n",
       "0  3.62160  8.6661 -2.8073 -0.44699      0\n",
       "1  4.54590  8.1674 -2.4586 -1.46210      0\n",
       "2  3.86600 -2.6383  1.9242  0.10645      0\n",
       "3  3.45660  9.5228 -4.0112 -3.59440      0\n",
       "4  0.32924 -4.4552  4.5718 -0.98880      0"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fetch_data(BANKNOTE_URL, BANKNOTE_PATH, 'banknote_data.txt')\n",
    "banknote_raw_data = load_data(BANKNOTE_PATH, 'banknote_data.txt')\n",
    "\n",
    "names = ['1', '2', '3', '4', 'label']\n",
    "\n",
    "banknote_raw_data.columns = names\n",
    "banknote_raw_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 1372 entries, 0 to 1371\n",
      "Data columns (total 5 columns):\n",
      "1        1372 non-null float64\n",
      "2        1372 non-null float64\n",
      "3        1372 non-null float64\n",
      "4        1372 non-null float64\n",
      "label    1372 non-null int64\n",
      "dtypes: float64(4), int64(1)\n",
      "memory usage: 53.7 KB\n"
     ]
    }
   ],
   "source": [
    "banknote_raw_data.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>1372.000000</td>\n",
       "      <td>1372.000000</td>\n",
       "      <td>1372.000000</td>\n",
       "      <td>1372.000000</td>\n",
       "      <td>1372.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.433735</td>\n",
       "      <td>1.922353</td>\n",
       "      <td>1.397627</td>\n",
       "      <td>-1.191657</td>\n",
       "      <td>0.444606</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>2.842763</td>\n",
       "      <td>5.869047</td>\n",
       "      <td>4.310030</td>\n",
       "      <td>2.101013</td>\n",
       "      <td>0.497103</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>-7.042100</td>\n",
       "      <td>-13.773100</td>\n",
       "      <td>-5.286100</td>\n",
       "      <td>-8.548200</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>-1.773000</td>\n",
       "      <td>-1.708200</td>\n",
       "      <td>-1.574975</td>\n",
       "      <td>-2.413450</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.496180</td>\n",
       "      <td>2.319650</td>\n",
       "      <td>0.616630</td>\n",
       "      <td>-0.586650</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>2.821475</td>\n",
       "      <td>6.814625</td>\n",
       "      <td>3.179250</td>\n",
       "      <td>0.394810</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>6.824800</td>\n",
       "      <td>12.951600</td>\n",
       "      <td>17.927400</td>\n",
       "      <td>2.449500</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 1            2            3            4        label\n",
       "count  1372.000000  1372.000000  1372.000000  1372.000000  1372.000000\n",
       "mean      0.433735     1.922353     1.397627    -1.191657     0.444606\n",
       "std       2.842763     5.869047     4.310030     2.101013     0.497103\n",
       "min      -7.042100   -13.773100    -5.286100    -8.548200     0.000000\n",
       "25%      -1.773000    -1.708200    -1.574975    -2.413450     0.000000\n",
       "50%       0.496180     2.319650     0.616630    -0.586650     0.000000\n",
       "75%       2.821475     6.814625     3.179250     0.394810     1.000000\n",
       "max       6.824800    12.951600    17.927400     2.449500     1.000000"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "banknote_raw_data.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The data set is complete (there are no missing values) and consists of 4 features, represented by 64-bit floating point numbers, alongside a binary label, indicating a class to which the example belongs to. The histograms are visualized in order to find basic information about the distribution of each feature's values. In the next step, the data is shuffled and split into training and test set, in a way to ensure equal distribution of each class' examples in both sets. The standarization of data set is achieved with StandardScaler class, which is a part of scikit-learn library."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[<matplotlib.axes._subplots.AxesSubplot object at 0x7ff442ded320>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x7ff442a25128>],\n",
       "       [<matplotlib.axes._subplots.AxesSubplot object at 0x7ff442cf76d8>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x7ff442c37c88>]],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "banknote_raw_data.hist(column = ['1', '2', '3', '4'], bins = 100, figsize = (15, 10))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0    0.555394\n",
      "1    0.444606\n",
      "Name: label, dtype: float64\n",
      "\n",
      "\n",
      "0    0.55515\n",
      "1    0.44485\n",
      "Name: label, dtype: float64\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import StratifiedShuffleSplit\n",
    "\n",
    "split = StratifiedShuffleSplit(n_splits = 1, test_size = 0.2, random_state = 42)\n",
    "for train_index, test_index in split.split(banknote_raw_data, banknote_raw_data['label']):\n",
    "    strat_train_set = banknote_raw_data.loc[train_index]\n",
    "    strat_test_set = banknote_raw_data.loc[test_index]\n",
    "    \n",
    "print(banknote_raw_data['label'].value_counts() / len(banknote_raw_data))\n",
    "print('\\n')\n",
    "print(strat_train_set['label'].value_counts() / len(strat_train_set))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = strat_train_set.copy()\n",
    "\n",
    "shuffle_indexes = np.random.permutation(len(train_data))\n",
    "train_data = train_data.iloc[shuffle_indexes]\n",
    "train_data = train_data.reset_index(drop = True)\n",
    "\n",
    "targets = train_data['label']\n",
    "data = train_data.drop('label', axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[<matplotlib.axes._subplots.AxesSubplot object at 0x7ff444d23128>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x7ff444cd12e8>],\n",
       "       [<matplotlib.axes._subplots.AxesSubplot object at 0x7ff444c7d6a0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x7ff444c9bb70>]],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x1080 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "scaler = StandardScaler()\n",
    "data_scaled = scaler.fit_transform(data)\n",
    "\n",
    "data_scaled = pd.DataFrame(data = data_scaled[:,:],\n",
    "                           index = None,\n",
    "                           columns = ['1', '2', '3', '4'])\n",
    "\n",
    "data_scaled.hist(bins = 100, figsize = (15, 10))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Stochastic gradient descent classification algorithm is implemented in scikit-learn as SGDClassifier class. The model was trained with the train set, which consists of 1097 examples from the input set.\n",
    "### Basic numeric metrics\n",
    "\n",
    "Firstly, the fit is evaluated with a confusion matrix, precision and recall score and F1 score."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SGDClassifier(alpha=0.0001, average=False, class_weight=None,\n",
       "              early_stopping=False, epsilon=0.1, eta0=0.0, fit_intercept=True,\n",
       "              l1_ratio=0.15, learning_rate='optimal', loss='hinge',\n",
       "              max_iter=1000, n_iter_no_change=5, n_jobs=None, penalty='l2',\n",
       "              power_t=0.5, random_state=42, shuffle=True, tol=0.001,\n",
       "              validation_fraction=0.1, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import SGDClassifier\n",
    "\n",
    "sgd_classifier = SGDClassifier(random_state = 42)\n",
    "sgd_classifier.fit(data_scaled, targets)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[602,   7],\n",
       "       [  7, 481]])"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.model_selection import cross_val_predict\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "data_predict = cross_val_predict(sgd_classifier, data_scaled, targets, cv = 3)\n",
    "confusion_matrix(targets, data_predict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision of predictions:  0.985655737704918\n",
      "Recall of predictions:  0.985655737704918\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import precision_score, recall_score\n",
    "\n",
    "print('Precision of predictions: ', precision_score(targets, data_predict))\n",
    "print('Recall of predictions: ', recall_score(targets, data_predict))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 score:  0.985655737704918\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import f1_score\n",
    "\n",
    "print('F1 score: ', f1_score(targets, data_predict))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Precision - recall curve\n",
    "\n",
    "The next metric used to evaluate the fit was a precision - recall curve. Firstly, the curve was plotted as a function of a threshold of a decision function. Based on tha graph, the threshold was calculated to ensure the optimal relationship between precision and recall scores. Of course, depending on the project's requirements, that value can be increased or lowered to put greater importance on one of the scores. A 3-fold cross - validation test was used.\n",
    "\n",
    "Secondly, the curve was plotted as precision in a function of recall."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Threshold for maximum values of precision and recall:  0.09451522975041904\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import precision_recall_curve\n",
    "\n",
    "data_pred_scores = cross_val_predict(sgd_classifier, data_scaled, targets, cv = 3, \n",
    "                                     method = 'decision_function')\n",
    "precs, recalls, thresh = precision_recall_curve(targets, data_pred_scores)\n",
    "\n",
    "plt.figure(figsize = (9, 6))\n",
    "plt.plot(thresh, precs[:-1], 'r-', label = \"Precision\")\n",
    "plt.plot(thresh, recalls[:-1], 'g-', label = \"Recall\")\n",
    "plt.xlabel('Threshold')\n",
    "plt.xlim([thresh[np.where(recalls == 1)], thresh[np.where(recalls == 0)[-1] - 1]])\n",
    "plt.legend(loc = 'center right')\n",
    "plt.grid(True)\n",
    "plt.show()\n",
    "\n",
    "print('Threshold for maximum values of precision and recall: ', thresh[np.where(recalls == precs)][0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize = (9, 6))\n",
    "plt.plot(recalls[:-1], precs[:-1], 'b')\n",
    "plt.xlabel(\"Recall\")\n",
    "plt.ylabel(\"Precision\")\n",
    "plt.xlim([0, 1.1])\n",
    "plt.ylim([precs[np.where(recalls == 1)][-1], 1.005])\n",
    "plt.grid(True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ROC Curve\n",
    "\n",
    "Lastly, the ROC curve was obtained. The area under curve was calculated with scikit - learn library's method. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjEAAAFzCAYAAADPF1gOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8li6FKAAAgAElEQVR4nO3deZzNdf//8cdrRNpdpVRIXaUyg7GMLdtkVyFLtrSNPdrXK6Uuvq2uq7oqxYhyVYqoUPZl7InIbCRJka5UTCVlmfP+/XEOv0mYg/nM55w5z/vtdm7O8jmf8/SOvOb9fp/Py5xziIiIiESbOL8DiIiIiBwLFTEiIiISlVTEiIiISFRSESMiIiJRSUWMiIiIRCUVMSIiIhKVTvA7wNEqXbq0u/DCCz07/2+//cYpp5zi2fnlyDT+/tHY+0vj7y+Nv38+/fTTH51zZx/Le6OuiLnwwgtZuXKlZ+dPS0sjOTnZs/PLkWn8/aOx95fG318af/+Y2dfH+l4tJ4mIiEhUUhEjIiIiUUlFjIiIiEQlFTEiIiISlVTEiIiISFRSESMiIiJRSUWMiIiIRCUVMSIiIhKVVMSIiIhIVFIRIyIiIlHJsyLGzMaY2TYzyzzM62ZmL5jZBjNLN7MaXmURERGRosfLmZjXgVZHeL01UDF06wO84mEWERERKWI8awDpnFtoZhce4ZB2wH+dcw742MxKmdl5zrnvvMpUaFJTYdw4v1NEpWo5OVCqlN8xYpLG3l8af39p/P3x3e7dx/V+P7tYlwU253m8JfTcX4oYM+tDcLaGMmXKkJaW5lmonTt3hn3+86ZOpczcuX95vtSaNQDkJCYWZLSYkJubS05Ojt8xYpLG3l8af39p/AvXr7m5DPnuO97cvv24zuNnERM251wqkAqQlJTkvGyXflTt2B97DDZtgmrV/vx848bQvTul+vQp4HRF31GNvxQojb2/NP7+0vgXnpkzZ9KnTx82b9/OwIEDefHFF4/5XH4WMd8C5fM8Lhd6LrLlXSpasCBYsHg4MyQiIlIUOOfo1asXY8aM4fLLL2fx4sVcccUVx1XE+PkV6ynAjaFvKdUFfo7Y/TCpqZCcHLz17RssXuDAjIuIiIgcmZlxzjnnMGjQIFavXs0VV1xx3Of0bCbGzN4GkoHSZrYFeBQoDuCcGwFMA64CNgC7gFu8ynJcUlODhQsEi5b9hYuWikRERI7of//7HwMHDmTgwIEkJyfz5JNPFuj5vfx2Urd8XnfAAK8+v8DsXzoaOVKFi4iISBicc4wdO5a77rqL33//ndatW3uy50hX7A1H48YqYERERMKwadMmWrZsyS233EKVKlVIT0+nZ8+ennyWipjD2b8PZv/+FxEREcnX1KlTWbZsGcOHDyctLY1LL73Us8+Kiq9Y+2LcOPjsM23eFRERycfatWv5+uuvadWqFbfeeisdOnSgbNmynn+uipgjqVZNX58WERE5jL179/LMM88wZMgQKlSowNq1aylWrFihFDCg5SQRERE5Bp9++ilJSUk8/PDDtG/fnsWLF1OsWLFCzaCZGBERETkqa9eupU6dOpxzzjl88MEHtGvXzpccmokRERGRsGzduhWASpUq8fLLL5Odne1bAQMqYkRERCQfv/zyCwMGDODiiy9m3bp1APTp04dSPnf+1nKSiIiIHNa0adPo168fW7Zs4Y477qB8+fL5v6mQqIgRERGRv3DOkZKSwuuvv058fDxLly6lbt26fsf6Ey0niYiIyF+YGeeffz6PPPIIq1atirgCBlTEiIiISMjWrVtp37498+fPB+Dxxx9nyJAhnHjiiT4nOzQVMSIiIjHOOcerr75KfHw8M2bM4Ouvv/Y7UlhUxIiIiMSwjRs30qxZM3r37k21atXIyMjg5ptv9jtWWFTEiIiIxLDp06ezYsUKRo4cybx587jkkkv8jhQ2fTtJREQkxmRlZfHNN9/QunVr+vfvT4cOHTjvvPP8jnXUNBMjIiISI/bs2cOQIUOoXr06d911F7m5ucTFxUVlAQMqYkRERGLCihUrSEpK4tFHH6VTp04sWrSo0Bs2FjQtJ4mIiBRxa9eupW7dupx33nlMmTKFNm3a+B2pQGgm5lBSU2HBAr9TiIiIHJctW7YAwYaNI0aMICsrq8gUMKAi5tDGjQv+2r27vzlERESOwc8//0zfvn255JJLWLt2LQC9e/fmjDPO8DlZwdJy0uE0bgx9+vidQkRE5Kh8+OGH9OvXj++++467776bChUq+B3JM5qJOch5U6dqKUlERKKOc44bb7yRNm3acOaZZ/Lxxx8zbNgwTj75ZL+jeUZFzEHKzJ0bvKOlJBERiSJmRoUKFfjnP//JypUrqVWrlt+RPKciJq/UVEqtWaOlJBERiQpbtmyhXbt2zJs3D4ChQ4cyePBgSpQo4XOywqEiJi9t6BURkSgQCARITU0lISGB2bNns3nzZr8j+UJFzEFyEhM1CyMiIhFrw4YNNG3alL59+5KUlERGRgY33XST37F8oW8niYiIRJGZM2eyatUqRo0aRc+ePTEzvyP5RkWMiIhIhMvIyOCbb77h6quvpn///nTs2JFzzz3X71i+03KSiIhIhNq9ezePPvooNWrU4J577jnQsFEFTJCKGBERkQj08ccfU6NGDYYMGULXrl1ZsmRJ1DdsLGhaThIREYkw2dnZ1K9fn7Jly/LRRx9x1VVX+R0pImkmRkREJELs/6p0fHw8o0aNIjMzUwXMEaiIERER8VlOTg69e/fmkksuITs7G4CUlBROP/10n5NFNi0niYiI+Gjy5Mn079+f77//nnvvvZeLLrrI70hRQ0WMiIiID5xz9OjRg3HjxlG1alWmTJlCUlKS37GiipaTREREfGBmXHzxxQwdOpQVK1aogDkGKmJEREQKyf4L1s2dOxeAIUOG8PDDD8dMw8aCpiJGRETEY4FAgFdeeYWEhATS0tLYunWr35GKBBUxIiIiHlq/fj3Jycnceuut1K1bl8zMTG644Qa/YxUJ2tgrIiLioTlz5pCRkcGYMWO4+eabY7phY0FTESMiIlLA1qxZw+bNm7nmmmvo168fnTp14pxzzvE7VpGj5SQREZECsnv3bh5++GGSkpK47777DjRsVAHjDRUxIiIiBWDZsmVUr16dxx9/nO7du7N48WI1bPSYlpNERESO0/6GjeXLl2f69Om0atXK70gxQTMxIiIix+jrr78Ggg0bx4wZQ2ZmpgqYQqQiRkRE5Cjt2LGDnj17cumllx5o2HjzzTdz2mmn+Zwstmg5SURE5Ci89957DBgwgB9++IH777+fv//9735HilkqYkRERMLgnKNbt26MHz+eatWqMW3aNKpXr+53rJim5SQREZEwmBmXX345TzzxBJ988okKmAigIkZEROQwvv76a1q3bs2cOXMAeOyxx/jHP/5B8eLFfU4moCJGRETkLwKBAC+99BIJCQksWrSI//3vf35HkkPQnhgREZE8Pv/8c3r27MmSJUto2bIlI0eOpEKFCn7HkkPwdCbGzFqZ2edmtsHMHjzE6xeY2XwzW21m6WZ2lZd5RERE8jNv3jyys7MZO3Ys06dPVwETwTybiTGzYsBwoDmwBVhhZlOcc9l5DnsYmOCce8XM4oFpwIVeZRIRETmUL774gl9//ZU2bdrQt29fOnXqxNlnn+13LMmHlzMxtYENzrmNzrk9wDtAu4OOccDpoftnAFs9zCMiIvInf/zxBw899BD9+vXjgQceONCwUQVMdPByT0xZYHOex1uAOgcd8xgwy8xuA04Bmh3qRGbWB+gDUKZMGdLS0go6KwDVcnLIzc317PySv507d2r8faKx95fGv/BlZGQwbNgwNm/eTLNmzbj99ttZtGiR37HkKPi9sbcb8Lpz7t9mVg94w8wqO+cCeQ9yzqUCqQBJSUkuOTnZmzSlSpGTk4Nn55d8paWlafx9orH3l8a/cGVnZ3PHHXdQoUIFZs2aRfHixTX+UcjL5aRvgfJ5HpcLPZdXT2ACgHNuGVASKO1hJhERiWFfffUVEGzYOHbsWDIyMmjevLnPqeRYeVnErAAqmtlFZlYC6ApMOeiYb4CmAGZWiWAR84OHmUREJAZt376dm266icsuu+xAw8YbbriBU0891edkcjw8W05yzu0zs4HATKAYMMY5l2VmQ4CVzrkpwD3AKDO7i+Am35udc86rTCIiEnsmTpzIgAED2L59Ow8++CAXX3yx35GkgHi6J8Y5N43g16bzPjc4z/1soL6XGUREJDY55+jatSsTJkygZs2azJo1i8TERL9jSQFS2wERESlS9k/omxkJCQk8/fTTfPzxxypgiiAVMSIiUmR89dVXtGzZ8kDDxsGDB3P//fdzwgl+fxlXvKAiRkREol5ubi4vvPAClStXZtmyZWzbts3vSFIIVJqKiEhUy87OplevXixbtozWrVszYsQILrjgAr9jSSFQESMiIlFt0aJFrF+/njfffJPu3btjZn5HkkKiIkZERKLOp59+ypYtW2jXrh29e/emY8eOlC6ta6XGGu2JERGRqPH777/zwAMPULt2bR566KEDDRtVwMQmFTEiIhIVFi5cSGJiIs888wwpKSksWbKEYsWK+R1LfKTlJBERiXjZ2dk0btyYiy66iDlz5tC0aVO/I0kE0EyMiIhErI0bNwLBho1vvvkmGRkZKmDkABUxIiIScX788UduuOEGLr/8crKysgC4/vrrOeWUU3xOJpFEy0kiIhIxnHNMmDCB2267jR07dvDQQw9xySWX+B1LIpSKGBERiQjOOa677jomTZpEUlISc+bMoWrVqn7HkgimIkZERHzlnMPMMDMSExOpW7cud955p/odSb60J0ZERHyzceNGmjdvzqxZswB45JFHuPfee1XASFhUxIiISKHLzc3lueeeo3LlyqxYsYLt27f7HUmikEpdEREpVJmZmfTs2ZNPPvmEa665hldeeYVy5cr5HUuikIoYEREpVEuWLGHjxo2MGzeOrl27qmGjHDMtJ4mIiOdWrFjB+++/D0Dv3r1Zt24d3bp1UwEjx0VFjIiIeGbXrl3ce++91K1bl0ceeYRAIEBcXBxnnXWW39GkCFARIyIinkhLS6Nq1ar8+9//pnfv3ixZsoS4OP2zIwVHe2JERKTAZWVlceWVV3LxxRczb948rrzySr8jSRGkklhERArMF198AUBCQgLjxo0jPT1dBYx4RkWMiIgctx9++IHu3bsTHx9/oGFjt27dOPnkk31OJkWZlpNEROSYOed4++23uf322/nll1945JFHqFixot+xJEaoiBERkWPinKNjx468//771KlTh9GjR5OQkOB3LIkhKmJEROSo5G3YWLNmTRo2bMjtt99OsWLF/I4mMUZ7YkREJGwbNmygSZMmzJw5E4BBgwZx1113qYARX6iIERGRfO3bt49//etfVKlShVWrVpGTk+N3JBEtJ4mIyJFlZGTQs2dPVqxYQdu2bXn55ZcpW7as37FEVMSIiMiRLVu2jE2bNvHOO+/QuXNn9TuSiKHlJBER+Yvly5czadIkAHr16sW6devo0qWLChiJKCpiRETkgN9++427776bevXq8dhjjx1o2HjmmWf6HU3kL1TEiIgIAHPnzqVKlSo899xz9O/fXw0bJeJpT4yIiJCVlUWzZs2oWLEiCxYsoFGjRn5HEsmXSmwRkRi2fv16INiw8Z133mHNmjUqYCRqqIgREYlB27Zto2vXriQkJJCZmQlAly5dOOmkk3xOJhK+fIsYMzvJzP5hZiNCjy8xs9beRxMRkYLmnOPNN9+kUqVKvP/++zz22GNcdtllfscSOSbh7IkZA2QADUKPtwLvAtO9CiUiIgXPOUf79u2ZPHky9erVY/To0VSqVMnvWCLHLJwipqJzrpuZXQfgnNtlulCAiEjUyNuwsXbt2jRp0oQBAwao35FEvXD2xOwxs5KAAzCzi4A9nqYSEZECsX79eho3bsyMGTMAeOihh9RxWoqMcIqYocAMoJyZjQXmAw95mkpERI7Lvn37ePrpp6latSoZGRns3LnT70giBS7f5STn3HQzWwlcARhwn3Num+fJRETkmKxZs4aUlBRWrVpF+/btGT58OOedd57fsUQKXL5FjJnNcs61ACYf4jkREYkwn3zyCd9++y0TJ06kY8eOfscR8cxhixgzKwGUBMqY2WkEZ2EATgcuKIRsIiISpqVLl7J161Y6depEr169uO666yhVqpTfsUQ8daQ9MQOALODy0K/7bzOBEd5HExGR/OzcuZM77riDBg0aMHToUAKBAGamAkZiwmGLGOfcc8658sADzrkLnHPlQ7cE59zzhZhRREQOYfbs2VSpUoUXXniBAQMGsHjxYjVslJgSzsbe583sciCe4PLS/ufHeRlMREQOLysrixYtWnDZZZexaNEiGjRokP+bRIqYcDb2Pgy0ILisNBNoCSwGVMSIiBSytWvXUqlSJRISEpgwYQJt2rShZMmS+b9RpAgKZ96xC3Al8J1z7gYgETjF01QiIvIn//vf/+jUqRNVq1YlKysLgOuuu04FjMS0cIqY351zucC+0LeU/gdU8DaWiIhAsGXA66+/Tnx8PB9++CFDhw7l0ksv9TuWSEQIp3fSajMrRbAR5ErgF+ATT1OJiAjOOdq2bcuHH35IgwYNePXVV9VxWiSPIxYxoUaPjznncoDhZjYTON05t6pQ0omIxKC8DRsbNGhAq1at6N+/v755JHKQI/6NcM45YHaexxuOpoAxs1Zm9rmZbTCzBw9zTGczyzazLDPTZmERiWnr1q2jYcOGTJ8+HYAHHniAAQMGqIAROYRw/lZ8ZmbVj/bEZlYMGA60Jvj17G5mFn/QMRWBfwD1nXMJwJ1H+zkiIkXBvn37eOKJJ0hMTCQ7O5tdu3b5HUkk4oWzJ6Y6sMLMvgR+I9h+wDnnauTzvtrABufcRgAzewdoB2TnOaY3MNw5t4PgSdVYUkRizurVq+nfvz8bNmygU6dOvPTSS5QpU8bvWCIRL5wipu0xnrsssDnP4y1AnYOOuRTAzJYAxQjuv5lx8InMrA/QB6BMmTKkpaUdY6Qjq5aTQ25urmfnl/zt3LlT4+8Tjb1/PvroI3766Sf++c9/0qhRI9auXcvatWv9jhVT9Oc/OoVzxd4vPf78ikAyUA5YaGZVQhuJ82ZIBVIBkpKSXHJysjdpSpUiJycHz84v+UpLS9P4+0RjX7gWL17M1q1b6dy5M40bN6Zx48Zcc801fseKWfrzH5283Cn2LVA+z+Nyoefy2gJMcc7tdc59BawnWNSIiBRJv/76KwMHDqRhw4Y88cQTBxo2nnrqqX5HE4k6XhYxK4CKZnaRmZUAugJTDjrmA4KzMJhZaYLLSxs9zCQi4psZM2ZQuXJlXn75Ze644w41bBQ5TuHsicHMygEVnXPzzexE4ATn3G9Heo9zbp+ZDSTYb6kYMMY5l2VmQ4CVzrkpoddamFk2kAvc55z76Xh+QyIikSgzM5PWrVtTqVIllixZQr169fyOJBL1wmkAmQIMBM4ALibYcuBloFl+73XOTQOmHfTc4Dz3HXB36CYiUqQ458jOziYhIYHKlSszadIkrr76ak488US/o4kUCeHMY94O1CXYbgDn3HrgHC9DiYhEu++++46OHTuSmJhIZmYmAB06dFABI1KAwili/nDO7dn/IHQRO/MukohI9HLO8dprrxEfH8/06dN58sknufzyy/2OJVIkhbMnZomZ3Q+UNLMrgQHAh97GEhGJPoFAgDZt2jBt2jQaNWrEqFGj1HFaxEPhFDH3E7zQ3DrgDoKbcUd6GUpEJJrsb9gYFxd34Hovffv21TePRDwWzt+wq4FXnXPtnXPXOudecc4FvA4mIhINsrOzueKKK5g2Lfgdhvvvv18dp0UKSTh/y64DNpjZa6Gu1MW8DiUiEun27NnD0KFDqV69Ol988QW7d+/2O5JIzMm3iHHO3UDwInRTgVuAjWY2wutgIiKRauXKldSqVYvBgwfToUMHsrOzad++vd+xRGJOWBe7c87tNrPJwO8EL1zXGejnZTARkUi1Zs0afvzxRyZPnkzbtsfaI1dEjle+MzFm1tzMXgW+BK4H/guc63UwEZFIsmDBAsaPHw9ASkoKa9euVQEj4rNw9sT0AWYAlZxzPZxzU/JeN0ZEpCj75Zdf6N+/P8nJyTz11FMHGjaefvrpfkcTiXnh7Im5zjk30Tn3e2EEEhGJFNOmTSMhIYHU1FTuvvtuNWwUiTCH3RNjZgucc43NbAfg8r5EsO3RmZ6nExHxSWZmJldffTXx8fFMnDiROnXq+B1JRA5ypI29V4Z+LV0YQURE/OacIzMzkypVqlC5cmXef/99WrdurX5HIhHqsPOieS5oN9o5l5v3BowunHgiIoXj22+/5dprr6V69eoHGjZee+21KmBEIlg4i7tV8z4IXeyuljdxREQKl3OOUaNGER8fz+zZs3n66aepVKmS37FEJAxH2hPzAPAgcJqZbd//NMH9MZqJEZGoFwgEuPrqq5kxYwbJycmMGjWKSy65xO9YIhKmI+2JeQb4N/AkwWIGgNBykohI1AoEAsTFxREXF0fTpk1p3749vXr10jePRKLMkf7GXuKc2we8ASTsv5lZVTOreoT3iYhErMzMTOrVq8eHH34IwL333kufPn1UwIhEoSPNxDwI9ASGH+I1BzTyJJGIiAf27NnDk08+yeOPP84ZZ5xBbq4mlUWi3WGLGOdcz9CvDQsvjohIwVuxYgUpKSlkZmbSvXt3nn/+ec4++2y/Y4nIcQqnd1IHMzstdP9BM5tgZoneRxMRKRiZmZns2LGDqVOn8tZbb6mAESkiwlkEfsw596uZXQFcBbwFjPQ2lojI8Zk/fz5vv/02ADfffDNr167lmmuu8TmViBSkcIqY/QvH1wAjnXOTAV39SUQi0s8//0yfPn1o0qQJw4YNO9Cw8bTTTvM7mogUsHCKmO/MbDjQFZhmZiXCfJ+ISKGaOnUq8fHxjB49mvvuu08NG0WKuCN9O2m/zgSXkV50zu0ws/PJc90YEZFIkJGRQdu2balSpQqTJ08mKSnJ70gi4rF8f0Rxzu0EsoBkM+sH/M05N93zZCIi+XDOsWbNGoADxcvKlStVwIjEiHC+nTQQeBe4IHSbYGa3eh1MRORINm/eTJs2bahZs+aBho1t27alRIkSPicTkcISznJSH6B2aEYGM3sCWAq87GUwEZFDCQQCjBo1ivvuu4/c3Fz+9a9/qWGjSIwKp4gxYE+ex3tDz4mIFKpAIECrVq2YPXs2TZs2JTU1lb///e9+xxIRn4RTxLwBLDezSQSLl2uBsZ6mEhHJI2/DxpYtW9KlSxdSUlIw089TIrEsnI29zwB9gV3ATqCfc+5fXgcTEQFIT0+nTp06Bxo23nPPPfTs2VMFjIiEfb2XP4DdeX4VEfHU7t27GTx4MDVr1uSbb77BOed3JBGJMOF8O2kQ8DZwHlAOGGdm//A6mIjEruXLl1OjRg2GDh1K9+7dyc7Opk2bNn7HEpEIE86emBuB6s65XQBm9jiwGnjSy2AiErvWrVvHr7/+yrRp02jdurXfcUQkQoXVdoA/FzsnhJ4TESkwc+fOZdy4cQDceOONZGdnq4ARkSMKp4jZDmSZ2atmNgrIAH40s2fN7Flv44lIUZeTk0OvXr1o1qwZzz777IGGjaeeeqrf0UQkwoWznPRR6Lbfxx5lEZEYM3nyZPr378+2bdt48MEHGTx4sBo2ikjY8i1inHOjCyOIiMSWjIwMrr32WhITE5k6dSo1a9b0O5KIRBn9yCMihcY5x+rVq4Fgw8apU6eyYsUKFTAickxUxIhIofjmm2+46qqrqFWr1oGGjddccw3Fixf3OZmIRKuwixgzO9HLICJSNAUCAYYPH05CQgKLFi3iueeeIz4+3u9YIlIEhHOxu9pmlgF8EXqcaGYvep5MRKJeIBCgRYsWDBw4kHr16pGZmcltt92mzbsiUiDC+T/JC8A1wE8Azrk1wJVehhKR6JabmwtAXFwcV199Na+99hozZ87kwgsv9DeYiBQp4RQxcc65rw96LteLMCIS/T777DNq167NlClTALjrrru4+eab1bBRRApcOEXMZjOrDTgzK2ZmdwLrPc4lIlHmjz/+YNCgQSQlJfHtt99qyUhEPBfOxe76E1xSugD4HpgTek5EBIBly5aRkpLCunXruOmmm3j22Wc588wz/Y4lIkVcOBe72wZ0LYQsIhKl1q9fz65du5gxYwYtW7b0O46IxIh8i5hQvyR38PPOuT6eJBKRqDBr1iy2bdtGjx49uPHGG+nUqROnnHKK37FEJIaEs2g9B5gbui0BzgF2exlKRCLX9u3bueWWW2jZsiUvvPDCgYaNKmBEpLCFs5w0Pu9jM3sDWOxZIhGJWJMmTWLAgAH8+OOPDBo0iIcfflgbeEXEN+Fs7D3YRUCZgg4iIpEtPT2dTp06Ub16dWbMmEG1atX8jiQiMS6cPTE7+P97YuKA7cCDXoYSkcjgnGPVqlXUrFmTqlWrMm3aNJo3b84JJxzLzz8iIgXriPPAFrw6VSJwduj2N+fc351zEwojnIj4Z9OmTbRq1YratWsfaNjYunVrFTAiEjGOWMQ45xwwzTmXG7r95VtKR2JmrczsczPbYGaHnb0xs45m5sws6WjOLyIFLxAI8OKLL1K5cmWWLl3Kiy++qIaNIhKRwvmR6jMzq+6cW300JzazYsBwoDmwBVhhZlOcc9kHHXcacAew/GjOLyIFLxAI0Lx5c+bNm0erVq0YMWIEFSpU8DuWiMghHXYmxsz2FzjVCRYgn5vZKjNbbWarwjh3bWCDc26jc24P8A7Q7hDHDQWeBv44yuwiUkDyNmxs27YtY8eOZdq0aSpgRCSiHWkm5hOgBtD2GM9dFtic5/EWoE7eA8ysBlDeOfeRmd13jJ8jIsdh1apVpKSkcN1115GcnMwdd9zhdyQRkbAcqYgxAOfcl158sJnFAc8CN4dxbB+gD0CZMmVIS0vzIhLVcnLIzc317PySv507d2r8C8nu3bsZO3Ys48ePp1SpUuzdu1dj7yP92feXxj86HamIOdvM7j7ci865Zx1J9HMAABhdSURBVPM597dA+TyPy4We2+80oDKQFvwSFOcCU8ysrXNu5UGflQqkAiQlJbnk5OR8PvoYlSpFTk4Onp1f8pWWlqbxLwRLliyhb9++rF+/np49ezJs2DDWrFmjsfeR/uz7S+MfnY5UxBQDTiU0I3MMVgAVzewigsVLV6D7/hedcz8Dpfc/NrM04N6DCxgRKXgbN25kz549zJ49m2bNmvkdR0TkmBypiPnOOTfkWE/snNtnZgOBmQQLojHOuSwzGwKsdM5NOdZzi8jRmz59Oj/++CM33HADPXr0oGPHjpx88sl+xxIROWb57ok5Hs65acC0g54bfJhjk4/380Tkr3766Sfuuusu3njjDWrVqsX1119PXFycChgRiXpHuthd00JLISIFzjnHu+++S3x8PG+//TaPPPIIixYtUsNGESkyDjsT45zbXphBRKRgZWRk0LlzZ2rWrMmsWbNITEz0O5KISIHSj2QiRYhzjhUrVgBQtWpVZsyYwccff6wCRkSKJBUxIkXEV199RYsWLahbt+6Bho0tW7ZUw0YRKbJUxIhEudzcXP7zn/9QuXJlli9fzvDhw9WwUURign5EE4ligUCAZs2akZaWxlVXXcWIESMoX758/m8UESkCVMSIRKF9+/ZxwgknEBcXR4cOHejduzfdunUjdPVrEZGYoOUkkSizcuVKatasyQcffADAbbfdRvfu3VXAiEjMUREjEiV27drF/fffT506dfjxxx8pWbKk35FERHyl5SSRKLBo0SJSUlLYsGEDvXv3ZtiwYZxxxhl+xxIR8ZWKGJEosGnTJgKBAHPnzqVJkyZ+xxERiQhaThKJUB999BH//e9/AejRoweZmZkqYERE8lARIxJhfvjhB66//nquueYaXn75ZQKBAGbGSSed5Hc0EZGIoiJGJEI453jnnXeIj4/n3Xff5dFHH2XhwoVq2CgichjaEyMSIdLT0+nWrRu1atVi9OjRVKlSxe9IIiIRTT/iifjIOcfy5csBSExMZNasWSxbtkwFjIhIGFTEiPjkyy+/pGnTplxxxRUHGjY2b96cYsWK+ZxMRCQ6qIgRKWS5ubk8++yzVKlShU8//ZQRI0aQkJDgdywRkaijPTEihSgQCNCkSRMWLlxImzZteOWVVyhbtqzfsUREopJmYkQKwb59+wCIi4ujc+fOvP3220yePFkFjIjIcVARI+KxTz75hOrVqx9o2DhgwAC6du2qho0iIsdJRYyIR3bt2sU999xDvXr12LFjByeffLLfkUREihTtiRHxwIIFC0hJSWHjxo3069ePp59+mtNPP93vWCIiRYqKGBEPbNmyhbi4ONLS0mjcuLHfcUREiiQtJ4kUkClTpvD6668D0L17dzIyMlTAiIh4SEWMyHHatm0bXbt2pV27dowcOfJAw8aSJUv6HU1EpEhTESNyjJxzvPXWW8THx/Pee+8xZMgQFixYoIaNIiKFRHtiRI5Reno6PXr0oG7duowePZr4+Hi/I4mIxBT9yChyFAKBAMuWLQOCDRvnzJnD4sWLVcCIiPhARYxImL744guaNGlCgwYNDjRsbNq0qRo2ioj4REWMSD727dvHsGHDqFq1Kp999hmjRo1Sw0YRkQigPTEiRxAIBEhOTmbJkiW0a9eOl19+mfPPP9/vWCIigmZiRA5p7969QLBhY7du3Rg/fjzvv/++ChgRkQiiIkbkIMuWLSMxMZH33nsPCDZs7Ny5sxo2iohEGBUxIiG//fYbd955J/Xr12fnzp3qdSQiEuG0J0YEmD9/PikpKWzatIkBAwbw5JNPctppp/kdS0REjkBFjAiwdetWSpQowcKFC2nYsKHfcUREJAxaTpKY9cEHH/Daa68BwYaN6enpKmBERKKIihiJOd9//z2dO3emffv2vPrqqwcaNp544ol+RxMRkaOgIkZihnOON954g/j4eCZPnszjjz9OWlqaGjaKiEQp7YmRmJGens6NN95IvXr1GD16NJUqVfI7koiIHAf9CCpFWiAQYPHixUCwYeO8efNYtGiRChgRkSJARYwUWZ9//jmNGzemUaNGBxo2XnnllWrYKCJSRKiIkSJn3759PPXUUyQmJpKVlcVrr72mho0iIkWQ9sRIkRIIBGjcuDFLly6lY8eOvPTSS5x77rl+xxIREQ9oJkaKhLwNG2+44QYmTpzIxIkTVcCIiBRhKmIk6i1ZsoQqVaowadIkAPr160fHjh19TiUiIl5TESNRa+fOndx+++00bNiQ33//nVKlSvkdSURECpGKGIlK8+bNo3Llyrz00ksMHDiQzMxMmjZt6ncsEREpRNrYK1Hp+++/p2TJkixatIj69ev7HUdERHygmRiJGpMmTeLVV18FoGvXrqxZs0YFjIhIDFMRIxHvu+++o2PHjnTq1ImxY8eqYaOIiAAqYiSCOed4/fXXiY+P56OPPuKpp55i/vz5atgoIiKAx0WMmbUys8/NbIOZPXiI1+82s2wzSzezuWZWwcs8El3S09O55ZZbqFKlCmvWrOGBBx7ghBO0jUtERII8K2LMrBgwHGgNxAPdzCz+oMNWA0nOuarAROAZr/JIdAgEAixatAgINmxcsGABaWlpXHbZZT4nExGRSOPlTExtYINzbqNzbg/wDtAu7wHOufnOuV2hhx8D5TzMIxFu7dq13H777TRu3PhAw8ZGjRpp+UhERA7Jy38dygKb8zzeEnrucHoC0z3MIxFq7969PPHEE1SrVo3NmzczduxYNWwUEZF8RcQGAzPrASQBjQ/zeh+gD0CZMmVIS0vzJEe1nBxyc3M9O7/8VSAQ4PbbbycrK4vk5GRSUlIoX748CxYs8DtazNm5c6f+7PtI4+8vjX908rKI+RYon+dxudBzf2JmzYBBQGPn3O5Dncg5lwqkAiQlJbnk5OQCDwtAqVLk5OTg2fnlgD179lC8eHHMjNtuu40yZcpw7bXXkpaWpvH3icbeXxp/f2n8o5OXy0krgIpmdpGZlQC6AlPyHmBm1YGRQFvn3DYPs0gEWbRo0Z8aNvbt25drr73W51QiIhJtPCtinHP7gIHATGAtMME5l2VmQ8ysbeiwYcCpwLtm9pmZTTnM6aQI+OWXXxgwYACNGjVi7969lC5d2u9IIiISxTzdE+OcmwZMO+i5wXnuN/Py8yVyzJ49m549e7JlyxbuvPNO/u///o9TTjnF71giIhLFImJjrxR9P/30E6eddhpLly6lbt26fscREZEiQEWMeMI5x8SJE8nJyaF379506dKFDh06UKJECb+jiYhIEaGriEmB27p1Kx06dKBz5868+eabOOcwMxUwIiJSoFTESIFxzjFmzBji4+OZMWMGzzzzDHPnzsXM/I4mIiJFkJaTpMCkp6fTq1cvGjZsyKuvvkrFihX9jiQiIkWYZmLkuOS9wnFiYiILFy5k/vz5KmBERMRzKmLkmGVlZVG/fn2aNGlCVlYWAA0aNFDDRhERKRT610aO2p49exg6dCjVq1fnyy+/5K233iI+Pt7vWCIiEmO0J0aOSiAQoFGjRixfvpxu3brxn//8h7PPPtvvWCIiEoM0EyNh2b17N8454uLi6NmzJ1OmTGHcuHEqYERExDcqYiRfCxYsoHLlykycOBGA3r1706ZNG59TiYhIrFMRI4f1yy+/0L9/f5KTkwkEApxzzjl+RxIRETlARYwc0qxZs0hISCA1NZW7776bjIwMGjdu7HcsERGRA7SxVw5px44dnHHGGUycOJE6der4HUdEROQvVMQIEGwZMH78eH7++Wf69u1L586d6dChA8WLF/c7moiIyCFpOUnYsmUL7dq1o1u3brzzzjsHGjaqgBERkUimIiaGBQIBUlNTSUhIYM6cOTz77LPMmTNHDRtFRCQqaDkphqWnp9OvXz+Sk5MZNWoUF198sd+RREREwqaZmBiTm5vLvHnzAKhWrRqLFy9m7ty5KmBERCTqqIiJIZmZmdSrV49mzZodaNh4xRVXaPlIRESikoqYGLBnzx4ee+wxatSowaZNm3j77bfVsFFERKKe9sQUcYFAgAYNGrBixQquv/56nn/+eUqXLu13LBERkeOmmZgi6o8//jjQsLFPnz58+OGHvPnmmypgRESkyFARUwTNmzePhIQE3n33XQB69erF1Vdf7XMqERGRgqUipgjJycmhd+/eNG3alLi4OM4//3y/I4mIiHhGRUwRMX36dBISEhgzZgz3338/6enpNGjQwO9YIiIintHG3iLi119/5ayzzmLy5MkkJSX5HUdERMRzKmKilHOOcePG8euvv9KvXz+uu+46OnTowAkn6D+piIjEBi0nRaHNmzfTpk0bevTowbvvvnugYaMKGBERiSUqYqJIIBBgxIgRJCQkMH/+fJ5//nlmzZqlK+6KiEhM0o/uUSQjI4Nbb72Vpk2bkpqaykUXXeR3JBEREd9oJibC7du3jzlz5gCQmJjI0qVLmTVrlgoYERGJeSpiItiaNWuoU6cOLVq0IDs7G4C6detq+UhERAQVMRFp9+7dPPLIIyQlJbFlyxYmTJhApUqV/I4lIiISUbQnJsIEAgHq16/Pp59+yo033sizzz7LWWed5XcsERGRiKMiJkL8/vvvlCxZkri4OG699VbOP/98WrVq5XcsERGRiKXlpAgwZ84c4uPjmTBhAgApKSkqYERERPKhIsZHO3bsoGfPnjRv3pwSJUpQrlw5vyOJiIhEDRUxPpk2bRrx8fGMHTuWBx98kDVr1lC/fn2/Y4mIiEQN7YnxyW+//ca5557LRx99RI0aNfyOIyIiEnVUxBQS5xxvvPEGO3fu5NZbb6VTp060b99e/Y5ERESOkZaTCsHXX39N69atuemmm3j//ffVsFFERKQAqIjxUCAQYPjw4VSuXJnFixfz4osvMnPmTF1xV0REpABoKsBD6enp3HbbbbRo0YKRI0dSoUIFvyOJiIgUGZqJKWB79+5l1qxZAFSrVo3ly5czffp0FTAiIiIFTEVMAVq9ejV16tShZcuWBxo21qpVS8tHIiIiHlARUwD++OMPBg0aRK1atdi6dSuTJk0iPj7e71giIiJFmvbEHKfc3Fzq16/PqlWruOWWW/j3v//N3/72N79jiYiIFHkqYo7Rrl27OOmkkyhWrBgDBw6kbNmytGjRwu9YIiIiMUPLScdg5syZVKpUifHjxwNwyy23qIAREREpZCpijsL27du56aabaNWqFSeffDIXXnih35FERERiloqYME2dOpVKlSoxbtw4Bg0axOrVq6lbt67fsURERGKW9sSEaffu3ZQrV46ZM2dSrVo1v+OIiIjEPE9nYsyslZl9bmYbzOzBQ7x+opmND72+3Mwu9DLP0XDO8dprrzF8+HAAOnXqxCeffKICRkREJEJ4VsSYWTFgONAaiAe6mdnBF0/pCexwzl0CPAc87VWeo7Fp0yZatmxJSkoKU6ZMwTkHQLFixXxOJiIiIvt5ORNTG9jgnNvonNsDvAO0O+iYdsDY0P2JQFPz8fK2uc4x8ocfqFy5MsuWLWP48OFMnz5dV9wVERGJQF4WMWWBzXkebwk9d8hjnHP7gJ+BszzMdESZv/3GP7ZupWHDhmRlZXHrrbcSF6e9zyIiIpEoKjb2mlkfoE/o4U4z+9zDjys9Y8aMH9Ww0TelgR/9DhGjNPb+0vj7S+Pvn8uO9Y1eFjHfAuXzPC4Xeu5Qx2wxsxOAM4CfDj6Rcy4VSPUo55+Y2UrnXFJhfJb8lcbfPxp7f2n8/aXx94+ZrTzW93q5VrICqGhmF5lZCaArMOWgY6YAN4XudwLmuf27aEVERESOwLOZGOfcPjMbCMwEigFjnHNZZjYEWOmcmwKMBt4wsw3AdoKFjoiIiEi+PN0T45ybBkw76LnBee7/AVznZYZjUCjLVnJYGn//aOz9pfH3l8bfP8c89qbVGxEREYlG+v6wiIiIRKWYLWKiuSVCtAtj7O82s2wzSzezuWam77sXoPzGP89xHc3MmZm+sVGAwhl/M+sc+juQZWbjCjtjURXG/3suMLP5ZrY69P+fq/zIWRSZ2Rgz22ZmmYd53czshdB/m3QzqxHWiZ1zMXcjuNH4S+DvQAlgDRB/0DG3AiNC97sC4/3OXRRuYY79lcDJofv9NfaFO/6h404DFgIfA0l+5y4qtzD//FcEVgN/Cz0+x+/cReEW5tinAv1D9+OBTX7nLio3oBFQA8g8zOtXAdMBA+oCy8M5b6zOxERdS4QiJN+xd87Nd87tCj38mOA1hqRghPNnH2AowV5mfxRmuBgQzvj3BoY753YAOOe2FXLGoiqcsXfA6aH7ZwBbCzFfkeacW0jwW8iH0w74rwv6GChlZufld95YLWKiriVCERLO2OfVk2B1LgUj3/EPTeOWd859VJjBYkQ4f/4vBS41syVm9rGZtSq0dEVbOGP/GNDDzLYQ/GbtbYUTTTj6fxuAKGk7ILHJzHoASUBjv7PECjOLA54FbvY5Siw7geCSUjLBWciFZlbFOZfja6rY0A143Tn3bzOrR/A6ZpWdcwG/g8mhxepMzNG0ROBILRHkqIUz9phZM2AQ0NY5t7uQssWC/Mb/NKAykGZmmwiuTU/R5t4CE86f/y3AFOfcXufcV8B6gkWNHJ9wxr4nMAHAObcMKEmwp5J4L6x/Gw4Wq0WMWiL4J9+xN7PqwEiCBYz2AxSsI46/c+5n51xp59yFzrkLCe5JauucO+beJvIn4fy/5wOCszCYWWmCy0sbCzNkERXO2H8DNAUws0oEi5gfCjVl7JoC3Bj6llJd4Gfn3Hf5vSkml5OcWiL4JsyxHwacCrwb2kv9jXOurW+hi5Awx188Eub4zwRamFk2kAvc55zTLPBxCnPs7wFGmdldBDf53qwfXguGmb1NsDgvHdpz9ChQHMA5N4LgHqSrgA3ALuCWsM6r/z4iIiISjWJ1OUlERESinIoYERERiUoqYkRERCQqqYgRERGRqKQiRkRERKKSihiRGGBmuWb2WZ7bhUc49sLDdZotbGaWZGYvhO4nm9kVeV7rZ2Y3FmKWaupqLBJZYvI6MSIx6HfnXDW/Qxyt0EX29l9oLxnYCSwNvTaioD/PzE4I9Uo7lGoE22BMK+jPFZFjo5kYkRgVmnFZZGarQrcrDnFMgpl9Epq9STeziqHne+R5fqSZFTvEezeZ2TNmlhE69pI8nzsvdL65ZnZB6PnrzCzTzNaY2cLQc8lm9mFo5qgfcFfoMxua2WNmdq+ZXW5mnxz0+8oI3a9pZgvM7FMzm3morrhm9rqZjTCz5cAzZlbbzJaZ2WozW2pml4Wu8DoE6BL6/C5mdoqZjQn93lab2aG6gYuIh1TEiMSGk/IsJb0fem4b0Nw5VwPoArxwiPf1A/4TmsVJAraELsfeBagfej4XuP4wn/uzc64K8BLwfOi5F4GxzrmqwFt5Pncw0NI5lwj86QrNzrlNwAjgOedcNefcojyvrQNKmNlFoae6AOPNrHjoszo552oCY4DHD5OzHHCFc+5uYB3Q0DlXPZTpCefcntD98aHPH0+wt9c851xt4EpgmJmdcpjzi4gHtJwkEhsOtZxUHHjJzPYXIpce4n3LgEFmVg54zzn3hZk1BWoCK0JtIU4iWBAdytt5fn0udL8e0CF0/w3gmdD9JcDrZjYBeO9ofnMEm/Z1AZ4K/doFuIxgM8vZoZzFgMP1YnnXOZcbun8GMDY06+QIXRr9EFoAbc3s3tDjksAFwNqjzC4ix0hFjEjsugv4HkgkOCv7x8EHOOfGhZZZrgammVlfwAjOpPwjjM9wh7n/1wOd62dmdUKf9amZ1QzvtwHAeIK9tt4Lnsp9YWZVgCznXL0w3v9bnvtDgfnOufahZay0w7zHgI7Ouc+PIqeIFCAtJ4nErjOA75xzAeAGgjMVf2Jmfwc2OudeACYDVYG5QCczOyd0zJlmVuEwn9Elz6/LQveX8v8bql4PLAqd52Ln3HLn3GCCnYPLH3SuX4HTDvUhzrkvCc4mPUKwoAH4HDjbzOqFzl/czBIOkzOvM4BvQ/dvPsLnzwRus9A0jwW7r4tIIVIRIxK7XgZuMrM1wOX8eTZiv85Appl9RnBp5r/OuWzgYWCWmaUDs4G/bJgN+VvomDsIzvwA3AbcEnr+htBrENxTkhH6evdSYM1B55oKtN+/sfcQnzUe6EFwaYnQPpZOwNOh3+NnwF82Lx/CM8CTZraaP89Wzwfi92/sJThjUxxIN7Os0GMRKUTqYi0injCzTUCSc+5Hv7OISNGkmRgRERGJSpqJERERkaikmRgRERGJSipiREREJCqpiBEREZGopCJGREREopKKGBEREYlKKmJEREQkKv0/F2D+ENxxCOcAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 648x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC score:  0.9991150502032357\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve, roc_auc_score\n",
    "\n",
    "false_positive_rates, true_positive_rates, thresh_roc = roc_curve(targets, data_pred_scores)\n",
    "\n",
    "plt.figure(figsize = (9,6))\n",
    "plt.plot(false_positive_rates, true_positive_rates, 'r')\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.xlabel('False positive rate')\n",
    "plt.ylabel('True positive rate')\n",
    "plt.xlim([-0.05, 1])\n",
    "plt.ylim([0, 1.05])\n",
    "plt.grid(True)\n",
    "plt.show()\n",
    "\n",
    "print('AUC score: ', roc_auc_score(targets, data_pred_scores))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, the evaluated SGD model was applied to the test set. A confusion matrix, precision and recall scores and F1 score was calculated."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix for the test set: \n",
      " [[150   3]\n",
      " [  8 114]]\n",
      "Precision of predictions for the test set:  0.9743589743589743\n",
      "Recall of predictions for the test set:  0.9344262295081968\n",
      "F1 score for the test set:  0.9539748953974896\n"
     ]
    }
   ],
   "source": [
    "test_data = strat_test_set.copy()\n",
    "\n",
    "shuffle_indexes_test = np.random.permutation(len(test_data))\n",
    "test_data = train_data.iloc[shuffle_indexes_test]\n",
    "test_data = train_data.reset_index(drop = True)\n",
    "\n",
    "data_test = strat_test_set.drop(\"label\", axis = 1)\n",
    "targets_test = strat_test_set[\"label\"]\n",
    "\n",
    "data_test_scaled = scaler.transform(data_test)\n",
    "\n",
    "data_test_scaled = pd.DataFrame(data = data_test_scaled[:,:],\n",
    "                                index = None,\n",
    "                                columns = ['1', '2', '3', '4'])\n",
    "\n",
    "data_test_predict = cross_val_predict(sgd_classifier, data_test_scaled, targets_test, cv = 3)\n",
    "\n",
    "print('Confusion matrix for the test set: \\n', confusion_matrix(targets_test, data_test_predict))  \n",
    "print('Precision of predictions for the test set: ', precision_score(targets_test, data_test_predict))\n",
    "print('Recall of predictions for the test set: ', recall_score(targets_test, data_test_predict))\n",
    "print('F1 score for the test set: ', f1_score(targets_test, data_test_predict))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The used model performed very well with the used data set, achieving **the F1 score over 98% for the train set and over 95% for the test set.** \n",
    "\n",
    "The practical use of binary classifier's metrics helps to understand the idea behind them and facilitates the further development in the machine learning field.    "
   ]
  }
 ],
 "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}