old/lr_poly_agg_tweets.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATAPATH = '../data/'\n",
    "XFERPATH = '~/downloads/'  # To transfer files from Google Colab\n",
    "FNAME = 'embeds_agg.csv'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "frompath = XFERPATH + FNAME\n",
    "topath = DATAPATH + FNAME\n",
    "!cp $frompath $topath"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.preprocessing import PolynomialFeatures\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.decomposition import PCA\n",
    "import pandas as pd\n",
    "import gender_guesser.detector as gender\n",
    "import numpy as np\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import f1_score\n",
    "from sklearn.metrics import roc_curve, auc\n",
    "import matplotlib.pyplot as plt\n",
    "from utils import get_rows, balance, balanced_split_by_time_and_id"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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></th>\n",
       "      <th>name</th>\n",
       "      <th>male</th>\n",
       "      <th>embed0</th>\n",
       "      <th>embed1</th>\n",
       "      <th>embed2</th>\n",
       "      <th>embed3</th>\n",
       "      <th>embed4</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th>time</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>425472153</th>\n",
       "      <th>2019-05-25 20:24:48+00:00</th>\n",
       "      <td>Queen K</td>\n",
       "      <td>False</td>\n",
       "      <td>9156</td>\n",
       "      <td>-31075</td>\n",
       "      <td>-21194</td>\n",
       "      <td>41023</td>\n",
       "      <td>33810</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1081873499621273601</th>\n",
       "      <th>2019-05-25 20:24:48+00:00</th>\n",
       "      <td>Alexandru Constantin</td>\n",
       "      <td>True</td>\n",
       "      <td>-34055</td>\n",
       "      <td>62140</td>\n",
       "      <td>-33476</td>\n",
       "      <td>1849</td>\n",
       "      <td>42774</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>987721418</th>\n",
       "      <th>2019-05-25 20:24:48+00:00</th>\n",
       "      <td>Susie Carmichael</td>\n",
       "      <td>False</td>\n",
       "      <td>-38618</td>\n",
       "      <td>12734</td>\n",
       "      <td>6150</td>\n",
       "      <td>6856</td>\n",
       "      <td>29976</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>980758839770451973</th>\n",
       "      <th>2019-05-25 20:24:52+00:00</th>\n",
       "      <td>Princess Sih.</td>\n",
       "      <td>False</td>\n",
       "      <td>-19252</td>\n",
       "      <td>-1482</td>\n",
       "      <td>-14205</td>\n",
       "      <td>-25178</td>\n",
       "      <td>-47939</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>838620336367218692</th>\n",
       "      <th>2019-05-25 20:24:53+00:00</th>\n",
       "      <td>Reid Kisselback</td>\n",
       "      <td>True</td>\n",
       "      <td>-40698</td>\n",
       "      <td>32644</td>\n",
       "      <td>3243</td>\n",
       "      <td>-35412</td>\n",
       "      <td>61766</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                               name   male  \\\n",
       "id                  time                                                     \n",
       "425472153           2019-05-25 20:24:48+00:00               Queen K  False   \n",
       "1081873499621273601 2019-05-25 20:24:48+00:00  Alexandru Constantin   True   \n",
       "987721418           2019-05-25 20:24:48+00:00      Susie Carmichael  False   \n",
       "980758839770451973  2019-05-25 20:24:52+00:00         Princess Sih.  False   \n",
       "838620336367218692  2019-05-25 20:24:53+00:00       Reid Kisselback   True   \n",
       "\n",
       "                                               embed0  embed1  embed2  embed3  \\\n",
       "id                  time                                                        \n",
       "425472153           2019-05-25 20:24:48+00:00    9156  -31075  -21194   41023   \n",
       "1081873499621273601 2019-05-25 20:24:48+00:00  -34055   62140  -33476    1849   \n",
       "987721418           2019-05-25 20:24:48+00:00  -38618   12734    6150    6856   \n",
       "980758839770451973  2019-05-25 20:24:52+00:00  -19252   -1482  -14205  -25178   \n",
       "838620336367218692  2019-05-25 20:24:53+00:00  -40698   32644    3243  -35412   \n",
       "\n",
       "                                               embed4  \n",
       "id                  time                               \n",
       "425472153           2019-05-25 20:24:48+00:00   33810  \n",
       "1081873499621273601 2019-05-25 20:24:48+00:00   42774  \n",
       "987721418           2019-05-25 20:24:48+00:00   29976  \n",
       "980758839770451973  2019-05-25 20:24:52+00:00  -47939  \n",
       "838620336367218692  2019-05-25 20:24:53+00:00   61766  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(DATAPATH+FNAME,index_col=[0,1],parse_dates=['time']).drop(['text'],axis=1)\n",
    "df.head().iloc[:,:7]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(29678, 514)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(Timestamp('2019-05-21 17:50:48+0000', tz='UTC'),\n",
       " Timestamp('2019-05-26 12:53:53+0000', tz='UTC'))"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(df.shape)\n",
    "df.index.get_level_values(1).min(), df.index.get_level_values(1).max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "xy_train, xy_valid, xy_test = balanced_split_by_time_and_id(df)\n",
    "\n",
    "X_train = xy_train.drop(['name','male'],axis=1).values / 1e6\n",
    "y_train = 1*xy_train.male.values\n",
    "\n",
    "X_valid = xy_valid.drop(['name','male'],axis=1).values / 1e6\n",
    "y_valid = 1*xy_valid.male.values\n",
    "\n",
    "X_test = xy_test.drop(['name','male'],axis=1).values / 1e6\n",
    "y_test = 1*xy_test.male.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.06617814 0.04185153 0.02776238 0.0250321  0.02302092 0.02204792\n",
      " 0.02121207 0.01935231 0.01814371 0.01740349 0.01651333 0.01546586\n",
      " 0.01518258 0.01496722 0.01415051 0.0137926  0.01365105 0.01320401\n",
      " 0.01297017 0.01280166 0.01199954 0.01167379 0.01158718 0.01104929\n",
      " 0.01079567 0.01038352 0.01028225 0.00996152 0.00982437 0.00939181\n",
      " 0.00929769 0.00905794 0.00875762 0.00873787 0.00849311 0.0084166\n",
      " 0.00828947 0.00810313 0.00796313 0.00784121 0.00769192 0.00762465\n",
      " 0.00748959 0.00730796 0.00713632 0.00708932 0.0068693  0.00682599\n",
      " 0.00674288 0.00654368 0.00649569 0.00634819 0.00625939 0.0061349\n",
      " 0.00606993 0.0058607  0.00583465 0.00576193 0.00563439 0.00557769\n",
      " 0.00537556 0.00520635 0.00510163 0.0050073 ]\n"
     ]
    }
   ],
   "source": [
    "pca = PCA(64)\n",
    "pca.fit(X_train)\n",
    "print(pca.explained_variance_ratio_) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/andy/anaconda/lib/python3.6/site-packages/sklearn/linear_model/logistic.py:433: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n"
     ]
    }
   ],
   "source": [
    "model = Pipeline([('pca',  PCA(50)),\n",
    "                  ('poly', PolynomialFeatures()),\n",
    "                  ('lr',   LogisticRegression(C=.5))])\n",
    "model = model.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6349206349206349"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train_pred = model.predict_proba(X_train)[:,1]\n",
    "f1_score(y_train, y_train_pred>.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1291  872]\n",
      " [ 809 1354]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.6169970380496697"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred = model.predict_proba(X_valid)[:,1]\n",
    "print( confusion_matrix(y_valid, (y_pred>.5)) )\n",
    "f1_score(y_valid, y_pred>.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1290  791]\n",
      " [ 811 1270]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.61323032351521"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test_pred = model.predict_proba(X_test)[:,1]\n",
    "print( confusion_matrix(y_test, (y_test_pred>.5)) )\n",
    "f1_score(y_test, y_test_pred>.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6150888995675157"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_test, y_test_pred>.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEWCAYAAAB42tAoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzs3Xd4FFX3wPHvIYHQI12poSqh9yJiQykKooiAiKAgAiK+YAML+rO/FkBeEVREVFRQAUUFQZAinSC9IzWhhRZqSDu/P2bBJYRkA9lMyvk8zz7Z2WlnJrt79t47c6+oKsYYY8zl5HA7AGOMMRmbJQpjjDHJskRhjDEmWZYojDHGJMsShTHGmGRZojDGGJMsSxTZiIh0FZFZbsfhNhEpKyKnRCQgHfcZIiIqIoHptU9/EpENInLLFaxn78FMSOw+CneIyC6gBBAPnAJ+B/qr6ik348qKPOe6l6rOdjGGEGAnkFNV49yKwxOLApVVdbuf9xNCBjlmc3WsROGutqqaH6gN1AGGuBzPFXHzV3JW+YWeGna+r1xmj98tligyAFU9AMzESRgAiEiQiLwvIntE5KCIjBGRPF7z7xGR1SJyQkT+EZFWnteDReRzEdkvIhEi8sb5KhYR6SEiCz3Px4jI+95xiMjPIjLI87ykiEwWkUgR2SkiA7yWe1VEfhSRCSJyAuiR+Jg8cXzlWX+3iLwkIjm84lgkIv8TkSgR2SwitydaN7ljWCQiw0XkKPCqiFQUkT9F5IiIHBaRb0TkGs/yXwNlgV881U3PJa4GEpF5IvK6Z7snRWSWiBT1iudhzzEcEZGXRWSXiLRI6n8pInlE5APP8lEistD7/wZ09fxPD4vIi17rNRSRJSJy3HPcH4lILq/5KiJPiMg2YJvntQ9FZK/nPbBSRG7yWj5ARF7wvDdOeuaXEZEFnkXWeM5HJ8/yd3veT8dFZLGI1PTa1i4ReV5E1gKnRSTQ+xx4Yg/zxHFQRIZ5Vj2/r+OefTXxfg961q0mIn+IyFHPui+k5ryKyC0iEp5oWe/YEr9XXxCRsyJS2Gv5Op7/R07P9KMisklEjonITBEpl1RM2Yqq2sOFB7ALaOF5XhpYB3zoNX8EMA0oDBQAfgHe9sxrCEQBd+Ak+1LADZ55PwGfAPmA4sBy4HHPvB7AQs/z5sBe/q1+LAScBUp6trkSGArkAioAO4CWnmVfBWKB9p5l8yRxfF8BP3tiDwG2Aj294ogDBgI5gU6e4yns4zHEAU8CgUAeoJLnXAQBxXC+oEYkda490yGAAoGe6XnAP0AVz/bmAe945oXiVA0285yL9z3H3uIy/9dRnvVLAQFAU09c5/f5mWcftYBzQFXPevWAxp5jCgE2Af/x2q4Cf+C8H/J4XnsIKOJZ52ngAJDbM+9ZnPfU9YB49lfEa1uVvLZdFzgENPLE3N1zzoK8zt9qoIzXvi+cU2AJ0M3zPD/QOKnznMR7sACw3xN7bs90o1Se11uA8GQ+W6+S6L0K/Ak85rX8e8AYz/P2wHagque8vgQsdvv7wu2H6wFk14fnzXwKOOn5MM0BrvHME+A0UNFr+SbATs/zT4DhSWyzhOfLJ4/Xa12AuZ7n3h9SAfYAzT3TjwF/ep43AvYk2vYQ4AvP81eBBckcW4AnjlCv1x4H5nnFsQ9PkvK8thzo5uMx7Lncvj3LtAdWJTrXKSWKl7zm9wN+9zwfCnznNS8vEEMSicLzRXQWqJXEvPP7LJ3omDtf5hj+A0z1mlbgthSO+9j5fQNbgHsus1ziRDEaeD3RMluAm73O36NJvH/PfxkvAP4PKHqZY75couji/X9K5riSO6+3kHKiWJBofi/+fa8Lzg+m85+DGXh+0Hjt+wxQLqU4s/LDqp7c1V5VC+C82W8Azld3FMP5QlrpqQo4jtPYXcwzvwzOL+DEyuH8Qt/vtd4nOL/KL6LOp2AizocV4EHgG6/tlDy/Dc92XsD5Ej9vbzLHVRTn1/dur9d24/waPC/CE4P3/JI+HsNF+xaR4iIy0VNNdQKYwL/n0lcHvJ6fwflljCemC/tT1TPAkctsoyjOL+Ok/jfJ7kdEqojIryJywHMMb3HpMSQ+7qc9VSRRnvMU7LXO5d4jSSkHPJ3o/10G59iT3HciPXFKY5tFZIWI3O3jfn2N0ZfzmpzEsf8INBGRkjglawX+8swrB3zodR6O4iSTUmRjligyAFWdD4zHqdYAOIzzC6qaql7jeQSr0/ANzhu/YhKb2ovza7yo13oFVbXaZXb9HXC/pw62ETDZazs7vbZxjaoWUNU23mEnc0iHcYr73nW7ZYEIr+lSIiKJ5u/z8RgS7/ttz2s1VbUgTpWMJLN8auzHqRoEnLpynOqepBwGokn6f5OS0cBmnKuRCuIkZkm0zIXj8LRHPA88ABRS1Wtwqu/Or3O590hS9gJvJvp/51XV75Lad2Kquk1Vu+Ak8/8CP4pIvuTWSWWMyZ3X0zg/qgCnbYZ/f1BdCDFRvMeBWTjn7kGcEuP5ZfbiVHN6n4s8qrrYhzizLEsUGccI4A4Rqa2qCTh12cNFpDiAiJQSkZaeZT8HHhGR20Ukh2feDaq6H+cD8IGIFPTMqygiNye1Q1VdBUQCY4GZng8QOFUiJzwNmHk8DaPVRaSBLweiqvHA98CbIlLAk4gG4fzSP684MEBEcopIR5w64empPQaPAjjVeMdFpBRO/by3gzjtLFfiR6CtiDQVp3H5/7j0CxwAz/9tHDBMnIsBAjwNuEE+7KcAcAI4JSI3AH19WD4O5/8XKCJDgYJe88cCr4tIZXHUFJHzCS7x+fgM6CMijTzL5hORu0SkgA9xIyIPiUgxz/Gffw/Fe2JL4PLn/lfgWhH5jzgXbxQQkUaJF0rhvG4FcnvizYnTpuDL+f4WeBjo4Hl+3hhgiIhU8xxbsOf9ma1ZosggVDUSpwH4Zc9Lz+M0qi31VEXMxmmYRFWXA48Aw3F+Rc7n31/vD+NU+2zEqbP+EbgumV1/B7TA68Pi+aJvi3MV1k6cX3Rjcao2fPUkzq+9HcBCz/bHec1fBlT2bPtN4H5VPV+lk9pj+D+cBtko4DdgSqL5bwMveaoTnknFMaCqGzzHMhGndHESp+H33GVWeQanEXkFTrXFf/Htc/YMzq/bkzhf3JNSWH4mTn36Vpxqu2gurmIZhpOsZ+EkoM9xGnLBqbf/0nM+HlDVMJw2qo9wzvd2kriSLRmtgA0icgr4EKfdJdpTTfcmsMizr8beK6nqSZyLENriVMltA269zD6SPK+qGoXTpjQWp8R6Ggi/zDa8TcN5/x1U1TVeMU31bHui53O3Hmjtw/ayNLvhzqQ7EemBcwNcM7djSS0RyY/zq7myqu50Ox5j0oOVKIxJgYi0FZG8nnr393F+2e5yNypj0o8lCmNSdg9OQ/s+nOqKzmpFcZONWNWTMcaYZFmJwhhjTLIyXQdZRYsW1ZCQELfDMMaYTGXlypWHVTXxPSY+yXSJIiQkhLCwMLfDMMaYTEVEdqe8VNKs6skYY0yyLFEYY4xJliUKY4wxybJEYYwxJlmWKIwxxiTLEoUxxphk+S1RiMg4ETkkIusvM19EZKSIbBeRtSJS11+xGGOMuXL+LFGMx+l++HJa4/SbUxnojTNwizHGmAzGbzfcqeoCEQlJZpF7gK88nastFZFrROQ6z8A1xhhjrsap/XBoFQf++pJx03Ne1abcvDO7FBcPtBLuee2SRCEivXFKHZQtWzZdgjPGmAxPFc4dh6gdcGIvnNwLB5bD/iXEH91BjgKlmLOuAVH5GgLfXPFu3EwUSQ0nmWRXtqr6KfApQP369a27W2NM9qUKf4+A/cthy0TntaBrIG8xuKYSFK3J6kIv8PjoaF4cejtd+1xPV+DdL1I1uONF3EwU4UAZr+nSOP39G2OM8aYJEH0MDq6Epa9DxEKo0Qs6zoGyt11YLCYmniFDZjNhwjreeus27r67Sprs3s1EMQ3oLyITgUZAlLVPGGOMR+Ra2DMHItfAP9OckkTO/HBtfeh/HIL+HcJeVYmIOEmpUgUoXjwf69f3pVixfGkWit8ShYh8B9wCFBWRcOAVICeAqo4BpgNtcAZyPwM84q9YjDEmw4s9A6v+55Qajm2DyNVQtAZUbAf3/Aylb0pytd27j/PkkzM4diyaBQt68PzzaT8UvT+veuqSwnwFnvDX/o0xJlOIWATL34Edv8I1FaHB89DweSAHFK0GAbkuu+oPP2ygb9/fGDiwMc880xSRpJp+r16mG4/CGGMyvSObndLDob+d0sP1D0DX5XBtA59WX7x4L1WqFKFevZIsW9aLihUL+zVcSxTGGONPsWfg7BHYMgl2/gZ75zmvFygDLcZAuTsgwLf7HI4ePcvzz//B9OnbmTz5ARo3Lu2/uL1YojDGmLQSe8a5p2Hn77BrBuxbDHHREJgX8hSFGj3h9o+h8A2Qymqi6Og46tb9hLZtq7BxYz+Cg3P76SAuZYnCGGOuVOxZOLoR9vzpNEJvmeQkhYJlocoD0ORVpzop8Mq/1DdvPsz06dsYNKgJy5c/RvHiaXc1k68sURhjjK+ObHKSQfgCOBXu3AkNkLsIVO4AbX+EyvelurSQlLNnY3nrrb8YPTqMV165GVV1JUmAJQpjjLm8mJMQtRNWDoPN30F8DJS5Ba5tCKEjnSqkHP75Gh0zJozNm4+wZk0fSpUq6Jd9+MoShTHGJLZ/OSz5P6fhOTAI8pWEpq9D7b6Qq4D/drv/JIMGzaJv3/o89VRjBg70z+WuqWWJwhiTvZ3YC+Hz4PgOOBvplByij0Lt/tBu8lW1L/gqPj6BMWPCePXV+fTuXZf69UuSI0fGSBJgicIYkx3FnoX9S2DpG7B3rtPgXOh6p3uM2z6C8q0h9zXpEsq5c3EArFixj/nzexAaWixd9psaliiMMdnH/mWw9jPY9iME5oGyLeCRLVA4bTrPS40TJ84xdOhcVq06wPz5PRg/vn26x+ArSxTGmKzp6FbY/K1zs9vZw0710ukD0HiokxzylXAttJkzt9Oz5zRatqzI5MkPuBaHryxRGGMyv5hTsPsPOL0ftv/kXMZ6KhxK3QTXNYKyt8JNb0O+65zGaZfs2nWca6/NT8GCQXz7bQeaNy/nWiypYYnCGJN5Hd4A6z+HNZ+AxkH5u6BIKDQcAiXqQZC7l5WeFxMTzwcfLOaDD5bw00+dadYsc43UaYnCGJP5bJsC0zo4z69tALd8AFW7+vXS1St1+nQMjRqNpWzZYFaseIzy5Qu5HVKqWaIwxmR8mgCLXoYTe5zR3U7sgtBucOfnPneol94OHz7D4sV7adfuer744h7q1y/pt27A/c0ShTEm4zl3Aha+CMe2OHdGH9/uvN74ZahwNxSv48qVSr5QVcaPX83gwXPo0aMW7dpdT4MGpdwO66pYojDGZBw7Z8DiV+DACsh3Ldw6EorVdIYAzVss2UF8Mop3313Ejz9uYsaMrtSte53b4aQJcQaayzzq16+vYWFhbodhjEkrCfGwZjSEfQBnDkGNXlCzt9MonUmqas6cieXNNxfw8MO1KF26ILlzBxIQkMPtsC4iIitVtf6VrGslCmOMO47vcPpSmtUTcuSE+k9DvYGQt7jbkaXKjBnbeOKJ6TRqVJrg4Nzky5fxSz2pZYnCGJN+NMHpcG/uAKd6qUQ9qDMAbh0OkrF+gadEVTl9OpbXXlvA6NF30bJlJbdD8htLFMYY/4o+Dn/0hj1zIPYUSCDUfQru+Qnyl3Q7ulSLj09g1KgVLFy4h++/78jixY9m2quZfGWJwhiTtuLOwTcNIPoIkMO5Qzpvcbh/NhSpmi69sfrLypX76N37VwoWDGL06LsAsnySAEsUxpi0dGIPfNcUTkVAt1WQu7BzxVKewm5HdlVOnjxH/vy52LHjGE891Yhu3WpmiwRxniUKY0zqJcTBsW1O30q7/4DD65xhQSPXQkgr6LYa8hZ1O8qrpqr88MNGBg6cyQ8/dKRjx2puh+QKSxTGmNQ5uhUm1IXY01CkmlOtVLGd0+FeiXpQKGs06p44cY5OnX4kPPwE339/P02blnE7JNdYojDG+ObQamcshzUfQ3B56LwI8meNG8q8nTsXx+bNh6lZswRdu9agU6dq5MwZ4HZYrspc16MZY9JXfAz8/SH80AK+rgORq+H2UfDo1iyZJObN20Xt2p8wYsQyRISHHqqZ7ZMEWInCGJOU7dPg7xHOMKEAtZ+AJq9C6WauhuVPw4YtYcSIpYwc2Zp77rne7XAyFOvCwxjjdMJ3YIXT19I/P8Hxf5xuNGo+DiXquh2d3yQkKF9+uZq2ba/n3Lk4ChYMokAB9wY28ifrwsMYc2XOHoHVH8Pioc50yaZOD60V20HuzDduQmqsX3+IPn1+JTY2gZtuKkelSpn7El5/skRhTHZ0+gBMbgWRa6BgObj1Q6jzZKbphO9qHT8eTevW3/Diizfx2GN1M1wHfhmNJQpjsouEeKcxeucMZxCggFzQaT6Ubu52ZOnmt9+2snjxXt5883a2b3+SoCD7CvSFX8+SiLQCPgQCgLGq+k6i+WWBL4FrPMsMVtXp/ozJmGzn7BGY8wRsmeRMFwmFW0c4/S1lE+HhJ3jqqd9Zu/YgH3/cBsCSRCr47UyJSAAwCrgDCAdWiMg0Vd3otdhLwPeqOlpEQoHpQIi/YjIm20iIc7rR2DoZ5j/tdKPRborT9pAj+1zumZCg5MghfPfdOmrUKM4339xH7tyWIFLLn2esIbBdVXcAiMhE4B7AO1EoUNDzPBjY58d4jMn6EuJh64/wW2fIEQh5r4X6z0KzNzPs2NL+snx5BH36/Mpnn7Xl2WdvdDucTM2fiaIUsNdrOhxolGiZV4FZIvIkkA9okdSGRKQ30BugbNmyaR6oMZneyXCYdp9ziStAhbbQ/uds0zjt7dSpGJ577g+mTt3M++/fkWWGI3WTPxNFUu/QxDdtdAHGq+oHItIE+FpEqqtqwkUrqX4KfArOfRR+idaYzOrPAbDqf0710j0/Qfk22a70AE4HfpGRZwgODqJIkTxs3NiPQoXyuB1WluDPRBEOePeiVZpLq5Z6Aq0AVHWJiOQGigKH/BiXMVnDlu/h107O89tHQe1+7sbjom3bjvDEE9MJDs7NDz905PXXb3M7pCzFnxcPrwAqi0h5EckFdAamJVpmD3A7gIhUBXIDkX6MyZjM78hm+LaJkyQq3A2DErJ1kvjss5U0afI5LVtW5LvvOrgdTpbktxKFqsaJSH9gJs6lr+NUdYOIvAaEqeo04GngMxEZiFMt1UMzW58ixqSX2LOw/C1YPdq5Oa7Vl1C4ittRuWbBgt00bFiKBg1K8fffj1O2bLDbIWVZfr1OzHNPxPRErw31er4RsMsRjElO7FnYMhFmPgqV74OH10CBUm5H5ZqDB0/xzDN/sGDBbn7/vSu1a1/rdkhZnl1QbExGFX0MIhbBT20Bgbr/gVuHux2VqyIjT1Oz5hi6d6/Fhg39yJ8/l9shZQuWKIzJaP751ZMcPKo8AC0/h1z53YvJZWvXHmTVqv10716blSt7U7p0wZRXMmnGesIyJqOIPg7fNXOSxM3vw6B4eFqh7aRsmyROnYrh2Wdn0aLFVyQkOM2XliTSn5UojHFb9DFYMwYWvggoPLIZCtvAOQCvvDKXyMgzrF/fj+LF87kdTrZlicKY9BZ7GtaNhcPrYfdsOLEL8peExi9Cjd5QsEyKm8jK9uyJYtCgmbz11u28++4d1gV4BmCJwpj0oAlw+iDM/Q9s/d55rfHL0OF3J0nkKuBufBlAbGw8H364jHfeWchTTzWiXLlgSxIZhCUKY/zt0Br4thHEn3PGgOjwO4S0dDuqDCU2Np4TJ86xYsU+li7tZaPNZTA+JQrPndVlVXW7n+MxJuuI2gWrR0HY+84ocp3mO3/NBceOnWXw4NlERZ1j4sT7mTTpfrdDMklIsVwnIncB64A/PNO1RWSqvwMzJtPaPQfGV4ex5Z37IO74BHrtsCSRyOTJGwkN/ZjAwByMGXO32+GYZPhSongNp3vwuQCqulpEKvk1KmMyG02A+c/C2k+cxupafaHjHMhXwu3IMpxdu45TrlwwQUGB/PxzZxo2zL53mWcWviSKWFU9Lhf3a2/9MRlz3ond8EUoJMQ6pYeK7SBPEbejynCio+N4++2/GDVqBX/99Qh33519+6nKbHxJFJtE5AEgh4iUB54Clvo3LGMygbhzsPojmP8MFK8DXZc7o8qZS+zff5LmzcdTq1YJVq/uYzfNZTK+vKv7A0OBBGAKTm+wQ/wZlDEZ3u458GMLyFUQGr8ETV6xJJGEAwdOsWXLYZo3L8cXX9xDs2Y2QmVm5MtFyi1V9XlVreN5DAZa+zswYzKcHdOdKqYPxEkSdZ+CJ6PgxtctSSQSH5/A6NErqFFjNEuWhCMiliQyMV/e3S/hlCS8vZjEa8ZkXeELYepdThXTQ39DsRqWHJLx9NOzCAvbx9y53alevbjb4ZirdNl3uoi0xBmmtJSIDPOaVRCnGsqY7GH2E7DmYwjtBneOdW6aM5c4efIcb7+9kEGDmvDKKzcTHJybHDkk5RVNhpdc1dMhYD0QDWzweszCqp5MdnBoDUyo7ySJ1l85D0sSl1BVpk7dRLVqH7N//ylEoFChPJYkspDLlihUdRWwSkS+UdXodIzJGHftXw7zBsG+Rc6Y1F2WQMnGbkeVYUVEnOT11xfw9df3cvPNIW6HY/zAl0rWUiLyJhAK5D7/oqraRdAma9m3FDZ941zyGtISem6Hayq6HVWGFBsbz/DhS4mIOMGHH7Zm5creJLrXymQhvlz1NB74AhCcKqfvgYl+jMmY9BW1y+nV9bsmcGwL3P2903GfJYkkLVq0h7p1P+XPP3cyYEAjAEsSWZwvJYq8qjpTRN5X1X+Al0TkL38HZozf7fwdVo2EnTOgUBW4fZTT9YZ96SXpzJlY8ubNyd9/7+fll5vTsWOoJYhswpdEcU6cd8M/ItIHiADsejeTOanC5olO9dK+xVDpXui6Aq6t73ZkGZaq8tVXaxg8eA4LFz7Ck082cjskk858SRQDgfzAAOBNIBh41J9BGZOmNAH2LYFj2+DPJyH2FFTuAA8uhevsSy85Bw6conPnHzl5MoZffulCxYo2TkR2lGKiUNVlnqcngW4AIlLan0EZkyZOH4CVw2HFu850SEuofB80exMK2Fs4OWfPxhIefoKyZYN5+OFadO9ey0aby8aSTRQi0gAoBSxU1cMiUg14HrgNsE+ayZhOH4S5T8GWSRAUDC3GQM3HQOyLzhczZ26nX7/p3H9/Vf773zt49NE6bodkXJbcndlvAx2ANTgN2FNxeo79L9AnfcIzJpU0Ab6u4ww7et8MKNfCutpIhSFDZjNp0gZGjWpD69aV3Q7HZBDJfYLuAWqp6lkRKQzs80xvSZ/QjEml8AUwpz+c3g+P7YGCZdyOKFOIj0/gq6/W0KVLDR59tA4vv3wzefPmdDssk4EkVxaPVtWzAKp6FNhsScJkSPGxMONhmHQzBF0DD6+1JOGjv//eT+PGn/PFF6s5evQslSsXsSRhLpFciaKCiJzvIVaAEK9pVPU+v0ZmjC/2/Ak/3O4877wQSja1+yB8tGPHMe6661vefvt2unevZfdEmMtKLlF0SDT9kT8DMSZVIhbDiv9CxEJo9AI0eB6CbNS0lKgqkydvYt++kwwY0Ijt258kXz7r6NAkL7lOAeekZyDG+CxyHUy8EWr1g957IGc+tyPKFHbsOEb//tPZvTuKMWPuArAkYXxil4OYzGXeIOfeiGsbwu0fWTWTD1QVEWHkyGU0b16OQYOakCtXgNthmUzErxeWi0grEdkiIttFZPBllnlARDaKyAYR+daf8ZhMTBWmdXCSxG3/gy6LLUn44K+/dlOv3qfs2RPFiBGtGDy4mSUJk2o+lyhEJEhVz6Vi+QBgFHAHEA6sEJFpqrrRa5nKwBDgRlU9JiLWh5S5VMQimNjMeW79Mvnk6NGzPPPMLGbN+ocPP2xFmTLWfmOuXIqJQkQaAp/j9PFUVkRqAb1U9ckUVm0IbFfVHZ7tTMS5N2Oj1zKPAaNU9RiAqh5K/SGYLCv8L/izP0SuheJ1oetyyGG/hpOjqhw/Ho2qUrx4PjZufIKCBYPcDstkcr5UPY0E7gaOAKjqGuBWH9YrBez1mg73vOatClBFRBaJyFIRaeXDdk1Wpwp/DoBJzSF3Yei1E7qttCSRgg0bDnHzzeMZOnQuRYrk5Z13WliSMGnCl0SRQ1V3J3ot3of1kqpA1kTTgUBl4BagCzBWRK65ZEMivUUkTETCIiMjfdi1ydT2zoVV/3Mue31gLgSHuB1Rhvfuu4u45ZYv6dy5OiNG2O8tk7Z8SRR7PdVPKiIBIvIfYKsP64UD3rfHlsbpBiTxMj+raqyq7gS24CSOi6jqp6paX1XrFytWzIddm0xr5+9Oo/Utw5xeXk2yFi3ag6rSsGEp1q7tQ79+DayXV5PmfHlH9QUGAWWBg0Bjz2spWQFUFpHyIpIL6AxMS7TMT3iqsUSkKE5V1A7fQjdZyplD8PdImNIaaveDegPdjihDi4g4QceOP9C9+0/s33+KW24J4brrCrgdlsmifEkUcaraWVWLeh6dVfVwSiupahzQH5gJbAK+V9UNIvKaiLTzLDYTOCIiG4G5wLOqeuQKj8VkVmHDYHQJWDMG6j9rJYkUbNlymNq1P6Fq1aKsW9eXkiUtQRj/EtXEzQaJFhD5B6dKaBIwRVVPpkdgl1O/fn0NCwtzMwSTlsI+gPnPQOuvIfQht6PJ0FasiODgwdPcdVdldu06TvnyhdwOyWQiIrJSVa/o2vIUSxSqWhF4A6gHrBORn0Sk85XszJgLFg2FD8RJErcMtySRjKioaPr3n07btt9x9mwsImJJwqQrn264U9XFwGIReRUYAXwDTPRjXCYr2zEdlr4Obb6BCndbZ34p6Nv3N/Lnz8XGjU9QuHAet8Mx2ZAvN9zlx7lRrjNQFfgZaOrnuExWtOwdWPkBnD0MFdpC1QfdjijD+uefo7z44p989FEbvvyyPTlz2j0kxj2+lCjWA78A76rqX36Ox2Q1x/+Bo5udK5p2z4LO17xBAAAgAElEQVS6T0GztyBnXrcjy5DOnYvjvfcWM2LEUp5//kaCg4MsSRjX+ZIoKqhqgt8jMVlLzEmY9RhsmQTF60DBctBxDpS9ze3IMqz4+AT27j3BqlUHWLmyN+XKXXLvqTGuuGyiEJEPVPVpYLKIXHJplI1wZ5I1tiLEnYH2v0DFu92OJkOLjDzNM8/8QZEieRg2rCWTJz/gdkjGXCS5EsUkz18b2c6kTtROOBvpjF1drIbb0WRoX365mueem023bjV59dVb3A7HmCQlN8Ldcs/Tqqp6UbIQkf6AjYBnLrXuc5jVCwpVtiSRjIiIE5Qq5VztNWvWQ9Sqda3LERlzeb7cmf1oEq/1TOtATBZwZKOTJKo+BN3Xux1NhnT6dAzPP/8Hdep8wqFDp+nevbYlCZPhJddG0QnnktjyIjLFa1YB4Li/AzOZ0PovoGAItPna7UgypM2bD9Oq1QSaNSvLunV9KV7cxvo2mUNybRTLccagKI0zUt15J4FV/gzKZEJz/wN/fwj3zXA7kgwnPPwEkZGnCQ0txpdftufmm0PcDsmYVLls1ZOq7lTV2araQFXneD2Wq2psegZpMrijW5wk0forKG9jIZwXF5fA8OFLqF17DMuWRRAUFGhJwmRKyVU9zVfVm0XkGBcPOCSAqmphv0dnMr5fOzv3SlR9CEK7uR1NhtK16xSOHDnD4sU9qVKliNvhGHPFkqt6Oj/cadH0CMRkMtHHYPpDsHM63Dcdyrd2O6IM4fjxaIYNW8ILL9zEyJGtKF48HyJJDfZoTOaRXNXT+buxywABqhoPNAEeB6wVLrtKiIcNX8KownBwJTy8xpIEoKp8++06QkNHERl5mtjYeEqUyG9JwmQJvnTh8RPQQEQqAl8BvwHfAna7bXaiCgeWw4LnIXw+hLSCDtZwfd7q1Qd4991FTJnSicaNS7sdjjFpypdEkaCqsSJyHzBCVUeKiF31lF2oQsQi+KWDM1xp0erQeSGUutHtyFx37lwc77yzkNy5A3n++Wb8/ffj5MhhJQiT9fiSKOJEpCPQDWjveS2n/0IyGUZCHPzQwilBlLwROv0Fhau4HVWG8OefO+nb9zdCQ4sxcqRzpZclCZNV+ZIoHgX64XQzvkNEygPf+Tcs47pzUc441iLQYyMUqep2RBnCuXNxBAUFMmfODt577w7atbve7ZCM8bsUx8wGEJFAoJJncruqxvk1qmTYmNl+pgp/DYYV7zrTj0dA/pLuxpQBJCQon322kjfe+Is1a/rYSHMm07maMbN9GeHuJuBrIALnHoprRaSbqi66kh2aDGznDJjTH6J2OONY1/uP2xFlCDt2HKNr1ymIwG+/PWhJwmQ7vlQ9DQfaqOpGABGpipM4rigzmQxq0Suw9DWo1Q/q9LeqJuDUqRiioqIJDg6iV686PPJIHWuHMNmSL73H5jqfJABUdROQy38hmXR3+qCTJKr3hBajLEkAP/+8mWrVPubbb9dRpEheevasa0nCZFu+lCj+FpFPcEoRAF2xTgGzhrNHnW7Bt0+F3IXh5vfcjihD6NVrGgsX7mH8+Hu49dbybodjjOtSbMwWkdzAAKAZThvFAuB/qhrt//AuZY3ZaWTPXJjSGnIEQuuvofK9bkfkqtjYeH74YSNdulRn7dqD3HBDUYKCfPkdZUzm4LfGbBGpAVQEpqrqu1eyA5MBndgNk1tCtR5wyzDIld/tiFy1ZMle+vT5jRIl8nHXXZVtICFjErlsG4WIvIDTfUdX4A8RSWqkO5PZbPoOPgtxhiq9fVS2TxJLl4bTocP3DBnSjJkzHyI4OLfbIRmT4SRXougK1FTV0yJSDJgOjEufsIxfaAJMfxCqPOCMHRGQPW+wV1UmTFhLUFAgHTuGsnlzfwoWDHI7LGMyrOSuejqnqqcBVDUyhWVNRrf8XRgW4Dxv8zUEZs8vxi1bDtOixdcMH76U8uWvQUQsSRiTguRKFBW8xsoWoKL32Nmqep9fIzNpY+tkCHsf9i+FhkOg4WAIyL5XNw8dOo927arwxBMNCQy03z7G+CK5RNEh0fRH/gzEpLGTETCxGZzY5VQ1NX0VQlq6HZUr/vjjH4YOncfMmQ8xcWIHGyPCmFS6bKJQ1TnpGYhJQ1G7YGx5yFMU+kdBUEG3I3LFwYOnGDhwJkuWhPPRR62tismYK2QXimc1sWedJFGoMjyyGST7Va/ExycQHR1HVNQ5ypULZuzYduTNmz0b7o1JC379FhGRViKyRUS2i8jgZJa7X0RURKz/qKtx7gSMKgQIPPR3tkwSq1btp2nTcXz44TKqVCnC22+3sCRhzFXy+ZtERFJVbheRAGAU0BoIBbqISGgSyxXAufN7WWq2b5Lw0TUQfw76HsyW90e88MIcWrX6hscfr8fgwc3cDseYLCPFRCEiDUVkHbDNM11LRP7nw7Yb4oxdsUNVY4CJwD1JLPc68C7gSpcgWYIqrBwBKPQ5AHmLuR1RulFVVqyIAKBx49KsX9+XRx+1Xl6NSUu+lChGAncDRwBUdQ1wqw/rlQL2ek2He167QETqAGVU9dfkNiQivUUkTETCIiMjfdh1NqIJMKMbzBsILUZDvhJuR5Rudu06Ttu239Gjx8+cOhVDu3bXU6xYPrfDMibL8SVR5FDV3Ylei/dhvaR+0l3ogVBEcuCMdfF0ShtS1U9Vtb6q1i9WLPv8Wk5RfKxzE92mb+CuiVCrj9sRpZvFi/dSv/6n3HhjGVatepz8+bPvvSHG+JsvVz3tFZGGgHraHZ4EtvqwXjhQxmu6NLDPa7oAUB2Y57mu/Vpgmoi0U1XrHjYl//wKf/R2nvc/DkHB7saTThYu3EOOHEK9etexYsVjlC9fyO2QjMnyfClR9AUGAWWBg0Bjz2spWQFUFpHyIpIL6AxMOz9TVaNUtaiqhqhqCLAUsCThiz1z4ae2EHIn9N6bLZLEkSNn6NVrGp07/0hUVDRBQYGWJIxJJymWKFT1EM6XfKqoapyI9AdmAgHAOFXdICKvAWGqOi35LZhLqMLXtSFyLTT9P2gy1O2I0s399/9AjRrF2bjxCbtxzph05svARZ/h1bZwnqr29ldQycnWAxdtnQy/3A8PLoXrGrkdjd9t2hTJu+8uZsyYu1CF3Lnt/lBjrtTVDFzkS9XTbGCO57EIKA6cu5Kdmau05FUnQWTxJHH2bCwvvfQnzZuPp27dawkMzGFJwhgX+VL1NMl7WkS+Bv7wW0QmaeF/weH18PAatyPxK1Vl5cr9bN16hDVr+lCyZAG3QzIm27uSn2nlgXJpHYhJxrJ3YOEQKHs7FKvpdjR+sW/fSQYOnEnDhiV5+ummNGtW1u2QjDEevtyZfUxEjnoex3FKEy/4PzSDKoyv5iSJxkPh/lluR+QXo0Ytp1atMVSuXJh+/Rq4HY4xJpFkSxTi3OBQC4jwvJSgKbV+m7Sz6CU4stHpuylvcbejSXOHDp2mePF8REfHMX9+D0JD7WZKYzKiZEsUnqQwVVXjPQ9LEull4Uuw7C2o2TvLJYkTJ84xYMAMGjT4jLNnY3n66aaWJIzJwHy56mm5iNT1eyTmXwsGw7I34aZ34JYRbkeTppYvjyA0dBRnzsTy99+9yZPHugA3JqO7bNWTiASqahzQDHhMRP4BTuP04aSqasnDHxa/Civ+C20mQNWubkeTZnbsOAZAxYqFmDjxfmusNiYTSa6NYjlQF2ifTrFkbyveg12zYM9saPRilkkSMTHxvP/+YoYNW8KoUW3o1Km6JQljMpnkEoUAqOo/6RRL9jXvaVg5zEkQzd+FEnXcjihNqCp33vk1+fPnIiysNyEh17gdkjHmCiSXKIqJyKDLzVTVYX6IJ/sJ+8BJEu2mQOV73Y4mTRw+fIbPP/+b5567kQkT7qNUqQJ4egg2xmRCyTVmBwD5cboDT+phrtaJPTD/Gaj6UJZIEgkJyrhxq6hW7WMOHDhFTEw8pUsXtCRhTCaXXIliv6q+lm6RZDfRx+D7W5znrb9yNZS0MmPGNsaMCeP337tSp851bodjjEkjKbZRGD84sAK+aQj5S0K3VZCJf3GfORPLG28sIDS0GF271qBVq0oEBPhy1bUxJrNI7hN9e7pFkZ38NcRJEiXqw2N7oHhttyO6YtOnb6NatY/ZufM4LVpUQEQsSRiTBV22RKGqR9MzkGxh62RY/g60GA3VH4UcAW5HdEViY+MJDMzBlCmb+OSTu7nzzopuh2SM8SP7+ZdeVJ1Bh6p2hVp9ICCX2xGlWlxcAiNGLKVGjdHExiYwdmw7SxLGZAM2Gkx62TPH+XvLcHfjuELr1x+ie/efCA4O4qefOpMrV+YsDRljUs8SRXrY8yf8eAdU7wl5M1fnd1FR0SQkKLlyBfCf/zTioYdq2uWuxmQzVvXkb/Ofgx9ud6qcWo51OxqfqSqTJq0nNPRjpk3bQpUqRejWrZYlCWOyIStR+Muyt2DZ2xB7Cm4fBbX7uR2Rz1SV9u0nsXPnMX74oSNNm5ZxOyRjjIssUaS1hHj4vBKc2OW0R9TsDTnzuh2VT86di+P337dzzz038PzzN9KgQUly5rS2CGOyO0sUaW1KGydJ9NoBweXdjsZn8+btom/f36hSpQht2lS2UoQx5gJLFGklIR6W/B/sngWtxmeqJPHTT5sZMGAGI0e2pn37G9wOxxiTwViiSAvbpsC0Ds7zxkOhWnd34/HB+Q78KlcuTOvWldi48Qny589893YYY/zPEsXVmtkT1o+DkJZw34xM0W/TunUH6dPnNxISlE8/vZugoECCgtyOyhiTUVmiuBoxJ50kUW8Q3PKB29H4RFXp338GDz9ck8ceq0eOHBk/sRlj3GX3UVyN9eOdv5kgSfzyyxZatpxAfLwyb153Hn+8viUJY4xPrERxpWJPw9wBUOMxtyNJ1t69UTz11O+sX3+I0aPvIjDQfhsYY1LHEkVqxcfChi/gj8ed6RYfuxvPZcTFJZCQoOzYcYxatUrw7bcdyJ3b/t3GmNSzb47UUIWvasLRzXBDF2jzTYZsvF62LJzHH/+Vp55qxCOP1OHmm0PcDskYk4lZokiNL2s4SSKD3kyXkKD07z+dqVM388EHd9KlS3W3QzLGZAF+rbAWkVYiskVEtovI4CTmDxKRjSKyVkTmiEg5f8ZzVZa+AUc2wEN/Z7gkoaqsX3+IHDmEJk1Ks3FjPx58sIZ14GeMSRN+SxQiEgCMAloDoUAXEQlNtNgqoL6q1gR+BN71VzxXZcnrsOhlaDkOStRxO5qLbNt2hDvvnEDPntOIj0+gW7daFCqUx+2wjDFZiD9LFA2B7aq6Q1VjgInAPd4LqOpcVT3jmVwKlPZjPFdm3xJYPBTqPQ3VH3E7mov88ssWmjT5nDZtKrFo0aM2XrUxxi/82UZRCtjrNR0ONEpm+Z7AjKRmiEhvoDdA2bJl0yq+lB1aDd81hRL1oemr6bffFMyevYMyZQrStGkZVq16nDJlgt0OyRiThfnzJ2hSFeSa5IIiDwH1gfeSmq+qn6pqfVWtX6xYOo0Qt3I4fF0HAoLgwaWQK3/67DcZBw+e4qGHptCr1zQiI89QpEheSxLGGL/zZ4kiHPDuq7o0sC/xQiLSAngRuFlVz/kxHt/EnIKf73GGL208FBq9ADncH5MhIUG57bavuPvuymzY0I98+awDP2NM+vBnolgBVBaR8kAE0Bl40HsBEakDfAK0UtVDfozFd7tmQvgCp4O/8q3cjoY1aw7wxRerGT68JcuX97IEYYxJd36relLVOKA/MBPYBHyvqhtE5DURaedZ7D0gP/CDiKwWkWn+iscnZw7BnH5Q8kbXk8SpUzE888ws7rjja6pVK4YqliSMMa7w6w13qjodmJ7otaFez1v4c/+p8ksn2Po9FCgDzd29SldV+fXXrURGnmH9+n4UL57P1XiMMdmb3ZkdHwMH/3aSRJtv4fqOkMOd07J793EGDPidBx+sTufOzsMYY9yWvS+8P7ACRgTBd02gWg+o2sWVJBEfn8B77y2iXr1PadCgpA1HaozJULJviWLZ27DwBajSEe76zrUrm44dO8s11+Tm+PFoli7tRaVKhV2JwxhjLif7JYqDK+G3LnBsG9z0X6j/tCtJ4ujRswwZMpulSyNYtepx3nzz9nSPwRhjfJG9qp6ObIIJ9SEwD/TYCA2fcyVJ/PHHP1Sr9jG5cgWwYEEPG2nOGJOhZZ8SxZnDMD4UilaHh9e4EsLmzYcpWjQv5csXYtq0zjRoUMqVOIwxJjWyR4lCE2DJq05Ddfd16b77s2djGTp0Ls2ajWPVqv1UqlTYkoQxJtPIHiWKv4bA6lFwxyfpvuu4uAQaNhzL9dcXYc2aPpQqVTDdYzDGmKuR9RPFwZWwZRLUeRJq9k633R44cIqff97M44/X5+efO1OhQqF027cxxqSlrF31lBDnNF6XbAo3vZ0uu4yPT+Djj1dQo8Zodu06jqpakjDGZGpZu0Txg6eHkDYTQNInJ44d+zfffbeeefO6U61a8XTZpzHG+FPWTRQJcbBvkTOWhJ+TxMmT53jllXm0bVuFnj3r8thj9eySV2NMlpF1q55+vMNJFtdU9tsuVJUpUzYRGvoxx45FU6NGCQIDc1iSMMZkKVmzRHFiD+ydB7d+CHn80yVGfHwCCQnK11+vZcKEe7n55hC/7MdkXrGxsYSHhxMdHe12KCYbyZ07N6VLlyZnzpxpts2slyhiz8Bn5SAoGGo+nvabj41n2LAlTJu2lYULH2Hq1E5pvg+TNYSHh1OgQAFCQkIQsVKm8T9V5ciRI4SHh1O+fPk0227WqnqKj4GRnrEbHt0OgUFpuvnlyyOoU+cT5s/fzddf32sffpOs6OhoihQpYu8Tk25EhCJFiqR5KTbrlCjORcEnpZ3n/Y87JYo0cuTIGfLly0VMTDyvvHIz998fah9+4xN7n5j05o/3XNYpUcx7GnLmg36H0yxJqCrjx6+mWrWPmTt3J82alaVjx2r24TfGZCtZJ1Hs/gOqdYc8RdJkczEx8dx221eMGrWC3357kNat/Xf1lDH+EhAQQO3atalevTpt27bl+PHjF+Zt2LCB2267jSpVqlC5cmVef/11VPXC/BkzZlC/fn2qVq3KDTfcwDPPPOPGISRr1apV9OrVy+0wkvX2229TqVIlrr/+embOnJnkMqrKiy++SJUqVahatSojR44EYN68eQQHB1O7dm1q167Na6+9BkBMTAzNmzcnLi4ufQ5CVTPVo169enqR+FjV1WNU30f14Gq9WmfOxOjs2f+oquoff/yjcXHxV71Nkz1t3LjR7RA0X758F54//PDD+sYbb6iq6pkzZ7RChQo6c+ZMVVU9ffq0tmrVSj/66CNVVV23bp1WqFBBN23apKqqsbGxOmrUqDSNLTY29qq3cf/99+vq1b5/7tNin6mxYcMGrVmzpkZHR+uOHTu0QoUKGhcXd8ly48aN027duml8vPN9c/DgQVVVnTt3rt51111JbvvVV1/VCRMmJDkvqfceEKZX+L2buUsUMSdheE74sz/c9hEUr3VVm/v99+1Urz6aL79cg6rSokUFAgIy9yky5rwmTZoQEREBwLfffsuNN97InXfeCUDevHn56KOPeOeddwB49913efHFF7nhBmdY3sDAQPr163fJNk+dOsUjjzxCjRo1qFmzJpMnTwYgf/78F5b58ccf6dGjBwA9evRg0KBB3HrrrTz77LOEhIRcVMqpVKkSBw8eJDIykg4dOtCgQQMaNGjAokWLLtn3yZMnWbt2LbVqOZ/75cuX07RpU+rUqUPTpk3ZsmULAOPHj6djx460bdv2wvG+9957NGjQgJo1a/LKK69c2Gb79u2pV68e1apV49NPP72Cs3yxn3/+mc6dOxMUFET58uWpVKkSy5cvv2S50aNHM3ToUHLkcL5vihdPuVeH9u3b880331x1jL7I3I3ZEQudv/+JgatsNxgzJoz33lvMqFFtaNWqUhoEZ0wiH/ihbetpTXkZID4+njlz5tCzZ0/AqXaqV6/eRctUrFiRU6dOceLECdavX8/TTz+d4nZff/11goODWbfO6b7/2LFjKa6zdetWZs+eTUBAAAkJCUydOpVHHnmEZcuWERISQokSJXjwwQcZOHAgzZo1Y8+ePbRs2ZJNmzZdtJ2wsDCqV69+YfqGG25gwYIFBAYGMnv2bF544YULiWvJkiWsXbuWwoULM2vWLLZt28by5ctRVdq1a8eCBQto3rw548aNo3Dhwpw9e5YGDRrQoUMHihS5uDp74MCBzJ0795Lj6ty5M4MHD77otYiICBo3bnxhunTp0heStbd//vmHSZMmMXXqVIoVK8bIkSOpXLnyhdhr1apFyZIlef/996lWrRoA1atXZ8WKFSme77SQuRPFse1wXZMrThLx8QmMHh1GixYV6NKlOt271yJPnrS7ScWYi/j4pZ6Wzp49S+3atdm1axf16tXjjjvuAJwq58tdlJGaizVmz57NxIkTL0wXKpRyB5gdO3YkIMAZWbJTp0689tprPPLII0ycOJFOnTpd2O7GjRsvrHPixAlOnjxJgQIFLry2f/9+ihUrdmE6KiqK7t27s23bNkSE2NjYC/PuuOMOChd2br6dNWsWs2bNok6dOoBTKtq2bRvNmzdn5MiRTJ06FYC9e/eybdu2SxLF8OHDfTs5cFGbz3lJnd9z586RO3duwsLCmDJlCo8++ih//fUXdevWZffu3eTPn5/p06fTvn17tm3bBjjtT7ly5brkvPhD5q5X2TIJCl9/RauuXLmPRo3G8uOPGwkIEIKDc1uSMFlOnjx5WL16Nbt37yYmJoZRo0YBUK1aNcLCwi5adseOHeTPn58CBQpQrVo1Vq5cmeL2L5dwvF9LfE1/vnz5Ljxv0qQJ27dvJzIykp9++on77rsPgISEBJYsWcLq1atZvXo1ERERl3wZ5smT56Jtv/zyy9x6662sX7+eX3755aJ53vtUVYYMGXJh29u3b6dnz57MmzeP2bNns2TJEtasWUOdOnWSvB9h4MCBFxqXvR/nq+28lS5dmr17916YDg8Pp2TJkkku16FDBwDuvfde1q5dC0DBggUvVOO1adOG2NhYDh8+fGG98wnG3zJvolg53On0r+SNqV41OjqOHj1+5sknGzJ3bncqV06bK6WMyaiCg4MZOXIk77//PrGxsXTt2pWFCxcye/ZswCl5DBgwgOeeew6AZ599lrfeeoutW7cCzhf3sGHDLtnunXfeyUcffXRh+nzVU4kSJdi0adOFqqXLERHuvfdeBg0aRNWqVS/8ek+83dWrV1+ybtWqVdm+ffuF6aioKEqVckaOHD9+/GX32bJlS8aNG8epU6cAp3ro0KFDREVFUahQIfLmzcvmzZtZunRpkusPHz78QpLxfiSudgJo164dEydO5Ny5c+zcuZNt27bRsGHDS5Zr3749f/75JwDz58+nSpUqABw4cOBCqWT58uUkJCRcOEdHjhyhWLFiadpVx+VkzkShCotfhdr9oaZvl8apKj/8sIGuXacQFBTAmjV96N69tt0TYbKNOnXqUKtWLSZOnEiePHn4+eefeeONN7j++uupUaMGDRo0oH///gDUrFmTESNG0KVLF6pWrUr16tXZv3//Jdt86aWXOHbsGNWrV6dWrVoX6u7feecd7r77bm677Tauu+66ZOPq1KkTEyZMuFDtBDBy5EjCwsKoWbMmoaGhjBkz5pL1brjhBqKiojh58iQAzz33HEOGDOHGG28kPj7+svu78847efDBB2nSpAk1atTg/vvv5+TJk7Rq1Yq4uDhq1qzJyy+/fFHbwpWqVq0aDzzwAKGhobRq1YpRo0ZdqHZr06YN+/btA2Dw4MFMnjyZGjVqMGTIEMaOHQs4FwKcP7cDBgxg4sSJF76z5s6dS5s2ba46Rl9IUnVoGVn9+vU1bOki+DC3c3OdD/dN7NhxjCeemM7evVGMGXM3zZqVTYdITXa3adMmqlat6nYYWdrw4cMpUKBAhr+Xwh/uu+8+3n77ba6//tLq96TeeyKyUlXrX8m+MmeJYscvzt8UkkRMTDyqyvLlEdx6awirVj1uScKYLKRv374EBaVtn26ZQUxMDO3bt08ySfhD5rzqKfooVGyX7CLz5++ib9/fGD68JZ07V092WWNM5pQ7d266devmdhjpLleuXDz88MPptr/MmSi2TYWceZOcdfZsLH37/sacOTsZObIVd95ZMZ2DM+ZfyV2Gaow/+KM5IXNWPe36HSrde9FLCQnK9u1HyZ07kAYNSrJxYz/uvbeqfUiNa3Lnzs2RI0f88sE1JinqGY8irS+ZzXwlCk1w/l7/wIWXNmw4RJ8+v1GwYBC//fYgTzxx6eVnxqS30qVLEx4eTmRkpNuhmGzk/Ah3aSnzJYqEWMgRCAG5ABg/fjXPPvsHr712C71710t+XWPSUc6cOdN0lDFj3OLXqicRaSUiW0Rku4hccjeKiASJyCTP/GUiEpLiRk/vh8C8TJ++jcOHz3DrrSGsW9eXvn0bWAd+xhjjB377ZhWRAGAU0BoIBbqISGiixXoCx1S1EjAc+G9K2405dYwOv7zMU0/9TkTECcqVu4Zrr82f0mrGGGOukD9/gjcEtqvqDlWNASYC9yRa5h7gS8/zH4HbJYXW582HSlC9ST3WretLrVrXpnnQxhhjLubPNopSwF6v6XCg0eWWUdU4EYkCigCHvRcSkd5Ab8/kuddeu229Z6Cn7K4oic5VNmbn4l92Lv5l5+JfV3x3nj8TRVIlg8TXCfqyDKr6KfApgIiEXelt6FmNnYt/2bn4l52Lf9m5+JeIhKW8VNL8WfUUDpTxmi4N7LvcMiISCAQDR/0YkzHGmFTyZ6JYAVQWkfIikgvoDExLtMw0oLvn+f3An2p3JxljTIbit6onT5tDf2AmEACMU9UNIvIaziDf04DPga9FZDtOSaKzD5u++oFssw47F/+yc2HwRjIAAAcPSURBVPEvOxf/snPxrys+F5mum3FjjDHpy+5QM8YYkyxLFMYYY5KVYROFX7r/yKR8OBeDRGTj/7d37yFylWccx78/q9akaloJitbLKl7amMYYY4kV1DQ2eMG0SsgqiXbFC8Za0Zr+IRFM0T/Eyx/GS1cbQhRUYsRL8IIVWY2ErBqq2cTgNQYRxIikoWgsNv31j/fdZtzMzpxds2dmZ54PDMycOee8zzzMnHfOe2aeV1KfpJclHdGIOMtQLxcV682WZEkt+9PIIrmQNCe/N96R9GjZMZalwGfkcEk9kt7Kn5Ny5hAtmaSlkrZI2jDI85K0OOepT9KUQju23XQ30sXvj4CjgL2BdcCEAetcDXTn+xcCyxsddwNzMR0Ym+/Pb+dc5PX2A1YBvcDURsfdwPfFMcBbwE/y4wMbHXcDc/EgMD/fnwBsbnTcI5SL04ApwIZBnj8HeIH0H7ZpwOtF9tusZxQjUv5jlKqbC9s9tr/OD3tJ/1lpRUXeFwC3ALcD35QZXMmK5OIK4D7bWwFsbyk5xrIUyYWB/fP9cez6n66WYHsVtf+L9lvgYSe9wI8lHVxvv83aUVQr//HTwdax/R+gv/xHqymSi0qXkb4xtKK6uZB0InCY7WfLDKwBirwvjgWOlbRaUq+ks0qLrlxFcrEImCfpU+B54I/lhNZ0hno8AZp3PordVv6jBRR+nZLmAVOB00c0osapmQtJe5CqEHeVFVADFXlf7EkafjqDdJb5mqSJtv85wrGVrUguLgKW2b5L0imk/29NtPtnQmsbwzpuNusZRZT/2KlILpB0JrAQmGX73yXFVrZ6udgPmAi8ImkzaQx2ZYte0C76GXnG9re2PwbeI3UcraZILi4DHgewvQbYh1QwsN0UOp4M1KwdRZT/2KluLvJwywOkTqJVx6GhTi5sb7M93naH7Q7S9ZpZtoddDK2JFfmMPE36oQOSxpOGojaVGmU5iuTiE2AGgKSfkzqKdpyjdiVwSf710zRgm+3P6m3UlENPHrnyH6NOwVzcAewLrMjX8z+xPathQY+QgrloCwVz8SIwU9JGYAfwZ9tfNi7qkVEwFzcAf5N0PWmopasVv1hKeow01Dg+X4+5GdgLwHY36frMOcCHwNfApYX224K5CiGEsBs169BTCCGEJhEdRQghhJqiowghhFBTdBQhhBBqio4ihBBCTdFRhKYjaYektytuHTXW7RisUuYQ23wlVx9dl0teHDeMfVwl6ZJ8v0vSIRXPLZE0YTfH+aakyQW2uU7S2O/bdmhf0VGEZrTd9uSK2+aS2p1r+wRSsck7hrqx7W7bD+eHXcAhFc9dbnvjbolyZ5z3UyzO64DoKMKwRUcRRoV85vCapH/k26+qrHO8pDfyWUifpGPy8nkVyx+Q9IM6za0Cjs7bzshzGKzPtf5/mJffpp1zgNyZly2StEDSbFLNrUdym2PymcBUSfMl3V4Rc5eke4YZ5xoqCrpJ+quktUpzT/wlL7uW1GH1SOrJy2ZKWpPzuELSvnXaCW0uOorQjMZUDDs9lZdtAX5jewrQCSyust1VwN22J5MO1J/mcg2dwKl5+Q5gbp32zwPWS9oHWAZ02v4FqZLBfEkHAOcDx9ueBNxaubHtJ4C1pG/+k21vr3j6CeCCisedwPJhxnkWqUxHv4W2pwKTgNMlTbK9mFTLZ7rt6bmUx03AmTmXa4E/1WkntLmmLOER2t72fLCstBdwbx6T30GqWzTQGmChpEOBJ21/IGkGcBLwZi5vMobU6VTziKTtwGZSGerjgI9tv5+ffwj4A3Avaa6LJZKeAwqXNLf9haRNuc7OB7mN1Xm/Q4nzR6RyFZUzlM2RdCXpc30waYKevgHbTsvLV+d29iblLYRBRUcRRovrgc+BE0hnwrtMSmT7UUmvA+cCL0q6nFRW+SHbNxZoY25lAUFJVec3ybWFfkkqMnchcA3w6yG8luXAHOBd4CnbVjpqF46TNIvbbcB9wAWSjgQWACfb3ippGanw3UACXrJ90RDiDW0uhp7CaDEO+CzPH3Ax6dv0d0g6CtiUh1tWkoZgXgZmSzowr3OAis8p/i7QIeno/Phi4NU8pj/O9vOkC8XVfnn0L1LZ82qeBH5HmiNheV42pDhtf0saQpqWh632B74Ctkk6CDh7kFh6gVP7X5OksZKqnZ2F8H/RUYTR4n7g95J6ScNOX1VZpxPYIOlt4GekKR83kg6of5fUB7xEGpapy/Y3pOqaKyStB/4LdJMOus/m/b1KOtsZaBnQ3X8xe8B+twIbgSNsv5GXDTnOfO3jLmCB7XWk+bHfAZaShrP6PQi8IKnH9hekX2Q9ltvpJeUqhEFF9dgQQgg1xRlFCCGEmqKjCCGEUFN0FCGEEGqKjiKEEEJN0VGEEEKoKTqKEEIINUVHEUIIoab/AUuk2yTpvNUWAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fpr, tpr, _ = roc_curve(y_test, y_test_pred)\n",
    "roc_auc = auc(fpr, tpr)\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(fpr, tpr, color='darkorange',\n",
    "         lw=1, label='ROC curve (area = %0.2f)' % roc_auc)\n",
    "plt.plot([0, 1], [0, 1], color='navy', lw=1, linestyle='--')\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.title('Receiver operating characteristic curve')\n",
    "plt.legend(loc=\"lower right\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}