DataExplor_Preproc_and_RF.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2154048, 18)"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from pylab import rcParams\n",
    "import seaborn as sb\n",
    "sb.set_style('darkgrid')\n",
    "rcParams['figure.figsize'] = 8,8\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "%matplotlib inline\n",
    "\n",
    "\n",
    "#import data\n",
    "train = pd.read_csv('Train.csv', index_col='user_id')\n",
    "#test=  pd.read_csv('Test.csv')\n",
    "#submission = pd.read_csv('SampleSubmission.csv')\n",
    "train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data exploration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(churn == 1) / (all_dataset): 18.75473527052322 %\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmsAAAGGCAYAAAAkfsYlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzs3XecVNX9//HXnbKVhV1hFUFRiHgEFGyI/mJXDMaoSWwECyrGqMGGsQAKChZEUVFRUEEFe0n8EoOxRE1iohFERUGOKNKLtAW278zc3x8zLLONGZgZ97K8n4/HPmbuvefc+9k7y/Lec5vjui4iIiIi4k2+5i5ARERERJqmsCYiIiLiYQprIiIiIh6msCYiIiLiYQprIiIiIh6msCYiIiLiYQprIiIiIh6msCYiIiLiYQprIiIiIh6msCYiIiLiYYHmLiBdIpGIGw6n79FZfr9DOteXbqovNaovNapvx3m5NlB9qVJ9qdmV6gsG/WuB4mTatpiwFg67lJSUp219hYV5aV1fuqm+1Ki+1Ki+Hefl2kD1pUr1pWZXqq+4uGBxsm11GFRERETEwxTWRERERDxMYU1ERETEw1rMOWsiIukSDofYsGENoVB12te9erWD63r3BGrVlxrVl5qWWF8gkEVRUTF+/45HLoU1EZF6NmxYQ05OHvn57XEcJ63r9vt9hMORtK4znVRfalRfalpafa7rUla2iQ0b1tCu3Z47vF0dBhURqScUqiY/v3Xag5qI7FocxyE/v3XKo/QKayIijVBQE5F0SMfvEoU1EREREQ9TWBMR8aCqqkoGD76ccDjcZJvBgy9n/vx5P2FV6bF48SL+8IdLOOGEo3jhhWlNtquqqmLo0Bu48MJzueSSASxfviwj9axcuYJ33vl7wnZr167h1ltvykgNicyfP4+HHrpvh/rW1NTwxz/+nlAo1GSb++67mzlzvmgw/5VXXuCtt97coe3WN3XqlKTa/elP17B58+a0bHN7XXHFpTvcd+TIoSxduiSN1WylsCYi4kFvvjmdY489Ab/fn7FtbCsIZlLr1q257ro/0b//Bdts98EH75Gf34pp015h/PiJtG7dJu21hEIhVq5cwXvvJQ5r7doVc+edY1Pe5o7s9wMO6M511924Q9sLBoMcdlhv3n//3SbbzJ37NT16HFRnXigU4m9/m07fvv12aLtbuK5LJBJh2rSnk2p///0PU1BQkNI2txVMt2XixOQCZWN+/euzeeGFqTvcf1t0NaiIyDZkvxwg58Vg2tbnOA4V/aupOm/b/5m8++7fGTnyztrp5557hnfeeQvH8XHkkf+PK6+8GoD333+PcePGsHlzKUOH3kavXocwY8ZfmT9/HkOG3AzATTddR//+F3DooYfTt+8xnHHGb5k161OGDLmZ0aNv49RTf8V//vMvQqEQd901lr333idt329jiop2o6hoN/7734+22S4YDLJmzRpc16V169bbbPvJJ//liScmEA5HKCwsZPz4x6moqODBB8fyww/fEwqFuPTSyznmmOOZMeOv/POf71NRUUEkEqG6uprFi3/g4osHcOqpp3HssScwevQIKisrALj++ps46KBerFy5gptuuo5p015hxoy/8tFH/6KyspIVK5Zx7LHHc9VV1zZZX/39np2dzaOPPkh5eTmFhYUMG3Y77dq145tv5jJmzGgcx0fv3n345JP/MG3aK8yePYuXXnqOsWMfYtOmjdxzzyhWrFhOdnYON900nP3268rkyZNYvXoVK1YsZ/Xq1Zx77u8455z+ABxzzPFMmvQop556WoPaFi36gb337tTgD4PZs2ex//4HEAhEo8KyZUu57757KCnZgN/vY/Toe+nYcS9eeGEq77//HjU11Rx77AkMGvQHVq5cwZAhg+ne/UCsnU/37j2oqqri4osH0LlzF0aOvJOhQ29g9erVVFdXc845/TnzzN8CcPbZp/PUU9OoqCjnT3+6hp49D+arr+ZQXFzMmDHjyM7OaXQfDx58OV27GubM+YKTT/4F/fqdxv33383q1asBuOaaIfTseTAbNmzgjjuGs3btWg488CBmzvwfkyc/R2FhIX37HsO77/4b13V57LGH+eST/+A4DgMHDuKkk05h9uxZPPnkRAoLC1m48HuM6caIEaNxHIdevQ7h7rvvIBQK1e6zdNnlwlpeZSn+stKE7fyb/RRUJ/7rJ5zfivKcVukoTUQEiB62WrFiOXvu2QGAjz/+Dx999C+eeOJZcnJy2LRpY23bcDjMk09O5eOPP2LKlCcZP/6xba67oqKC7t0P5Oqrr6+d16ZNG6ZMeZ4///lVXnhhGjfffGudPkuWLGLEiGGNru+RRyZRUFDAiBFDWbKk4aMOzztvAKee+qukv/d4HTp05Ntv5zNx4qO14bQxGzZsYOzYu3j00Sfo0KFj7f6ZOnUKhx3Wm2HDRrJ582Z+//uBHH54HwC+/dby7LMv0rp1mzpBCKCyspIHH5xAdnY2S5cu4fbbhzN5csPDtQsWfMvTTz9PMBhkwICzOOus8+jQoUOjNcbv91AoxODBl3PPPeMoKiriH/94hyeemMCwYSO5++47uPnmWznwwJ48/vgjja5r8uRJdO1quOeecXz22UzuvHMkzzzzAgBLlizm4YcnUl5ezoABZ/Gb35xNIBCgS5efNXnI/JNP/kOfPkc1mP/VV19izAG103fccSsXXHAxxx13AlVVVbiuy6effsLSpUt58slncV2XW24ZwhdfzGaPPdqzbNlShg+/gwMPjI7YffDBe7V1AgwdOoLWrdtQVVXJZZddxPHHn8huu+1Wp4Zly5Zy++13cfPNt3Lbbbfw4Yfv84tf/LLR7wOi/3a2fFa33z6cc889n169DmbVqlXccMNgnn/+NZ5++gkOO6w3F154CZ988l/efPP/Gqznn/98nwULLM888yIbN5Zw2WUX0avXoQAsWGCZNu0V2rUr5sorBzFnzpf06nUwPp+Pjh334rvvFnDAAd2arHFH7HJhzV9WSnhy4mFONydIuLIm8foGXQoKayItVtV5oYSjYNsjmfs0bdxYQqtWW3+vzJr1Kb/85enk5ERHFOIPBx533AkAGNONVatWJLF9P8cff2Kdeccdd2LtOv71rw8a9OnUad86/8k2ZtSoexJue3tUVVVy99138MILrzN69AheeeUFzj13ADfccA1XXDGYLl32q207d+5X9Op1CB06dAS27p9PP/2Ejz76Jy+++BwA1dVVrF69CoDevfs0eVg1FArx4IP3smDBt/h8fpYubfx524cf3rv2c9p33y6sWrWqybAWv9+XLFnEwoXfc/31fwQgEgnTtm07Nm/eTHl5OQce2BOAvn378d///rvBuubM+aL2cOxhh/Vm06aNlMUGIY466udkZWWRlZVFUVER69evY/fd98Dv9xMIBCkrKyMnJ7fO+j799BOGDRvZYDtr165ln332BaC8vIy1a9fU/rxlZ2fX9p058xMuueR8ACoqylm2bAl77NGe9u33rA1qjXn11Zf4178+BODHH1ezdOnSBmFtzz070LWrAcCYA1i5cts/4yed1Lf2/axZn7Jo0Q+102VlZZSXlzNnzpfcfXf0/L8jj/x/FBQ0HLXdMjrn9/vZbbe2HHLIocyfP5eCggK6devB7rvvAUDXrvuzatUKevU6GIiOGq9duwZQWBMRadGysrKprk7uvkxZWVkA+Hz+2nOh/H4/kcjWu6xXVVXXaV//cFcwmBXr52v0fKrmGFn7/vvvaNOmkKKiIu66ayzXXXcVjuNj06aNdO78s6TW4boud901lk6d9q0zf968r2uDb2Nefvl5iora8swzLxKJRDjppJ832i4Y3Hp4PLrvmg718fvddaFz5y5MmlT3HK50nFS/5bME8Pnqfp41NdVkZ2fVaV9ZWcnmzZtp1664wbqysxP/HLquywUXXMyvf31WnfkrV67Y5j6ePXsWs2Z9yqRJT5OTk8PgwZdTXV3VyPezdR9Hf8YbtomXm7s1iLpuhEmTnq4Nlumy5d9ctKa6+7i6uirt2wNdYCAi4jmtW7cmEolQVRX9j6l37z7MmPFXKisrAeocBm1M+/Yd+O67b4lEIqxevYpvvpmbUj1bRtYa+9pyIvioUfc0unxHD4HutVcnlixZzMKF35Obm8stt9zGhAkPccwxxze4b1WPHgfx5Zefs2LFcmDr/unT5yhee+3l2scDffvt/Ea3lZeXT3l5ee10WVkpbdu2w+fz8fbbM9J+IUanTvtQUrKBr7+eA0RH8hYu/J6CggLy8vKYO/drAP7xj3ca7d+r1yG8+270gojZs2fRpk0b8vO3fYRn48YS2rQpJBCoe/7l7NmzOPTQwxvts++++7Js2VIguo+Ki3evHQmrrq6msrKSPn2O4m9/m167/9as+ZENG9Y3uj6/P1B74n9ZWSkFBa3Jyclh8eJFzJv39Tbr3xG9ex/J66+/XDu9YIEF4KCDetVebPHpp5+wefOmBn179TqE999/l3A4zIYNG/jii8/p1q1Hwm0uXbqELl2S+2Nie2hkTUTEg3r37sOcOV/Qu3cfjjzy/7FgwbdcdtmFBAJBjjrq5/zhD39ssm/Pnr3Yc88OXHDBOeyzT2f239/8hJUntm7dWi677CLKysrw+RxeffVFnnvulTqBo3Xr1gwffjt33jkC13XJz2/FiBF3MmnSo/TseTAHHdSrtm1RURE33jiM4cNvJBJxKSoq4qGHHuPiiwcxfvw4Bg7sTyTi0qFDh9rz0uLtt19XfD4fAwf+jl/+8lf85jfncOutN/H3v/+NPn2OqjNakw7BYJA777yXhx66n9LSUsLhMOee+zu6dPkZt9wygrFj78RxfBx88KF1Dodvcemll3PPPaMYOLA/2dk5DB9+R8Jtzp49i6OOOrrB/E8++Q/HH39So32OPPLnjB49onb6tttGcd99dzN58kT8/gCjR4/hiCOOZNGiH7jiiksAyM3NY8SI0fh8DceCzjjjNwwc2J/99z+AoUNH8MYbf+b888+mU6d96N79wITfw/a67robeeCBexk4sD/hcJhevQ7hxhuHcemlv+f224fz9tszOPDAnrRt25a8vLw6fY899gS+/vorLr74dziOw1VXXUPbtu1YtqzpW3OsX7+O7Oxs2rZtl/bvxfHyA1O3R01N2C0pKU/YrmDdqqTOWcvJCVKZ5Dlrm9u2T6rGdCoszCOZ77e5qL7UqL7UpFrfqlWLad8+M1dEJvtsQWvn88orz3PbbaMzUkdTWtqzGX9qqdZXXl5eGxymTXuGdevWct11f0q5rmHDbuSKKwbTuXPnOvVdeun5PPHEs01evTh06J+46qpr2HvvTinXkIyf4vOtrq7G5/MRCAT4+us53H//mITnZCZT38svP09+fj6/+tWvGyxr7HdKcXHBZ0Djw5r1aGRNRMSDjDmAQw45nHA4nNF7rYm3fPzxR0yb9gzhcIj27fdk2LDbU15nTU0NxxxzHJ06NfwDZMqU57fZ98orB7Nu3dqfLKz9FFavXsWIEbcQibgEg0Fuvnl4WtbbqlXBNq9UTYXCmoiIR/3qV2c2dwmynQYNuqjBSfm33TaKn/1svyZ61HXSSadw0kmnpLWmYDC4w+cOduq0b4MLNJrbuHH38tVXX9aZd845/TnttDOS6r/33p14+unkRtK2R7Lb3xEZDWvGmH7AeMAPPGWtHdNEu7OA14De1tpZsXlDgUFAGLjGWvt2JmsVERFJ1eTJUz19mLYluOGGm5u7hJ9cxq4GNcb4gQnAqUB34HfGmO6NtCsArgX+FzevO9Af6AH0Ax6LrU9ERERkl5LJW3ccAXxnrV1ora0GXgIaG9MfDdwLVMbNOxN4yVpbZa39Afgutj4RERGRXUomw1pHYGnc9LLYvFrGmEOBva21f9veviIiIiK7gma7wMAY4wMeAC5Ox/r8fofCwrzE7Tb7cXMSP5TZ53PISaKdk+VParvp5vf7mmW7yVJ9qVF9qUm1vtWrHfz+zP0tm8l1p4PqS43qS01LrM9xkssoTclkWFsO7B03vVds3hYFwIHAh8YYgPbAdGPMGUn0bSAcdpO6r1JBdTipZ34mfZ+16jCbm+F+Uy39PleZpvpS09Lrc123zknivuwg6Txl3AdEqhL/ftlRCxZY1q5d0+hNUOPNnz+Pv//9b1x33Y21836q+5h99NE/+eGHH7jwwou3q19Lv89apqm+1Oxofa7bMKMUFxck3T+TYW0m0NUY05lo0OoPDNiy0Fq7Eai9za8x5kPgT9baWcaYCuAFY8wDQAegK/BpBmsVEWlSBLho8v8StkvW1EF90rau+kKhEAsWfMv8+fMShrUDDujOAQc0uO5ru7iui+u6jd6xfluOPvo4jj76uJS2LbKryFhYs9aGjDGDgbeJ3rpjirV2rjFmFDDLWjt9G33nGmNeAeYBIeCP1tr0PpxNRMTj3nrrTV566TnAYb/99mPw4CHcf//drF69GoBrrhlCz54HM3nyJFasWMaKFcvZfff2fPXVl1RXVzFnzpdceOHF7LlnB8aPHxd7yHQOw4aNoFOnfZk9exYvvfQcY8c+xOTJk1i9ehUrV65g1apVnHvu7zjnnP6N1rVy5QqGDBlM9+4HYu187r9/PEuWLGby5EnU1FTTocNeDBs2kry8PD7++CMeeeRBcnJy6dmzFytWLGfs2IeYMeOvzJ8/jyFDbmblyhXcc88oNm4sobCwiKFDR9K+fXvuuut28vPzmT//G9atW8dVV13NySen9x5kIjuDjJ6zZq2dAcyoN29EE22Przd9F3BX2ovy+SCQ+C4gVeFIUu3Yzr8mRUSSsXDh9zz77BQmTpxCYWEhmzZt5IEHxnLuuefTq9fBrFq1ihtuGMzzz78GwA8//MDjjz9FdnZOnSAE0YdmT5jwJIFAgJkz/8ekSRO46677GmxzyZLFTJjwBJs3lzJgwFn85jdnN/kYomXLljJ8+B0ceOBBlJSU8Oyzk3noocfIzc3lueee4eWXn2fAgIu47757ePTRJ+jQoSMjRw5rdF0PPngfp576K0499Ve8+eb/MX78fdxzzzgA1q5dy2OPPcXixYu45ZYhCmuyS9oln2Dwl9nLErbx+RwikcTPTT37knRUJCJS1+zZMznhhJMoLCwEoHXrNsya9SmLFv1Q26asrIzy8uh5MEcffSzZ2TmNrqu0tJQ777ydZcuW4DgOoVCo0XZHHfVzsrKyKCwspKioiPXr17H77ns02rZ9+z058MCDAJg79ysWLVrIlVcOAiAUqqFHj4NYsmQRHTp0pEOH6MX8ffv+gunT/9JgXXPnzuHuu6PhsV+/03j88Ydrlx177PH4fD46d+7C+vXrm9pdIi3aLhnWRER2Rq4bYdKkp8nOzm6wLCcnt8l+Tz01kUMPPZx77rmflStXcPXVf2i0XTCYVfve5/MRDjd99klOztZg6Louhx/ehzvuuLtOmwULbJP9kxUMxl+Vn/gPaJGWSMfwREQ86NBDe/PBB/9g48YSADZt2kjv3kfy+usv17ZpKgzl5eXVjrhBdGStuLgYgBkz/pr2Wnv0OIivvvqSZcuit8esqKhgyZLFdOq0DytWLGflyhUA/OMf7zba/8ADe/Lee9EnCr7zzlv07HlI2msU2ZlpZE1EJAEf6b2C0wcJbwXSpcvPGDjwUgYPvhyfz8/++xuuu+5GHnjgXgYO7E84HKZXr0O48caG54EdeujhPPfcs1x88QAuvPBizj//Iu6883aefXZywitEd0RRURHDh9/O7bcPp6Ym+hDz3//+Sjp12ochQ27mhhuuJicnl27dGr/y9Prrb+Luu+/gxRen1V5gICJbOa7bMoaVa2rCblL3WdvwI69dPTphu6TPWXvkNjYX7Z5UjenU0u9zlWmqLzUtvb5VqxbTvv0+aaxoq5Z6H6mmlJeXk5eXh+u6jBt3L3vvvTfnnXe+Z+pLN9WXmpZaX2O/U4qLCz4DDk+mv0bWREQkY/7617/w1lt/IxSqoWtXw5lnntXcJYnsdBTWRESkURs3lnDttVc1mD9+/GO0aVOY1DrOO+/8lEbSRERhTUREmtCmTSHPPPNCc5chssvT1aAiIiIiHqawJiIiIuJhCmsiIiIiHqawJiKSQF5lKQXrVqXlK3/NCvIqS3eojrvuup0PPngvzd9d5m3cWMLVV/+Bvn2P4YEH7m2yneu63HvvXVxwwTlcdNF5fP31nIzUs3nzZv7851eTanvFFZdmpIZE1q5dw6233tQs2xbv0QUGIiIJ+MtKCU+ekpZ1+XwO/ksugZxWaVnf9giHw/j9/p98u1lZ2Vx22ZX88MN3LFz4fZPt5sz5gmXLljBt2itUV1dRVlaW9lpCoRClpZv5y19e5be/PSdh+4kTU//cd2S/t2tXzJ13jk1529IyKKyJiHjQW2+9yUsvPQc47Lffftx2W/Rm3l9++Tkvv/wC69at46qrruaEE05m9uxZvPTSc4wd+xAADzxwLwcc0J1f/vJ0zj77dE48sS+zZv2PAQMu4o03Xqd79wP5/PNZbN5cytCht9GrV2Yf75Sbm0uvXgezfPnSbbYLBIJs2LCeUChEdnZOkw+mB/jmm7mMHz+OiooKsrKCjB//ONnZOUyc+Ciff/4ZNTXV/OY35/DrX5/F7NmzeOqpiRQUFLB48WKMMSxfvpyLLx5A7959uOSS3zN06A1s3ryJUCjE739/JcccczwAffsew7vv/pvZs2cxZcoTFBYWsnDh9xjTjREjRuM4TqP11d/v3br1YNy4eykp2UBOTg4333wr++yzL8uXL+OOO26lsrKCo48+jldffZF33/03K1eu4KabrmPatFeoqqpi3LgxzJ8/D7/fz9VXD+HQQw9nxoy/8tFH/6KyspIVK5Zx7LHHc9VV1+7w5yTepbAmIuIxCxd+z7PPTmHixCkUFhayadPG2mVr167lsceeYvHiRdxyyxBOOOHkhOtr06YNU6Y8D8Abb7xOOBzmySen8vHHHzFlypOMH/9Ynfbl5WVcddXvG13XyJF30rlzFx5+eByzZ3/WYPlJJ53ChRdevB3f7Va77bYb5eXl3HXX7YwceWeTQaimpoYRI4YxatTddOvWg7KyUrKysnnzzf8jPz+fp56aSnV1NVdeOYgjjjgSgG+/nc/UqS/ToUNHVq5cwcKF39feliQUCnH33feRn9+KkpIS/vCHizn66OMabH/BAsu0aa/Qrl0xV145iDlzvqRXr4Ob/H7i9/u1117Jn/40lL337sTcuV8zbtwYHn54IuPH38855/Snb99+vPHGa42uZ8sh26lTX2bx4kVcf/0fefHFP8dq+pann36eYDDIgAFncdZZ57HHHu23Y6/LzkBhTUTEY2bPnskJJ5xEYWH0xrOtW7epXXbsscfj8/no3LkL69evT2p9J510Sp3p4447AQBjurFq1YoG7fPy8hPeX+2aa25Iatvb49Zbb+bRR59g6tQpPPzwA1x77Q2MG3cvRx75//j5z4+pbbdkyWLatWtLt249AMjPjx5SnjnzE7777js+/PB9AMrKSlm2bCmBQIBu3XrQoUPHJrc9adIEvvzycxzHx5o1a1i/fh1t27ar06Zbtx7svvseAHTtuj+rVq3YZljbst/Ly8v56qs53HbbLbXLtjxD9euvv+Luu+8HoG/ffkyYML7BeubM+YKzzz4PgH322Zf27fdk6dIlABx+eG9atYp+//vu24VVq1YprLVACmsiIjuRYDAYNxV9frHfHyAS2fq8wurq6jp9cnJy60xnZWUB4PP5CYfDDbbRHCNrGzasZ+PGEjp06MiNNw7j1ltvYsqUJ5g/fy5XXXVNUutwXZfrr7+RPn2OqjN/9uxZ5ObmNtEL3nnnLUpKSpg8+TkCgQBnn316g30IW/cbgM/na3Tfxduy3103QkFBq4zcYDj+5yH63MpQ2rchzU9hTUTEYw49tDfDht1I//7n06ZN9DBo/Ohafe3bt2fRoh+orq6mqqqKzz6bSc+eTY/4JNIcI2uFhUW4rsvs2bM49NDDuemm4QwYcDYnnnhyg6DVqdM+rF27jm++mUu3bj0oLy8jKyubI444ijfeeI3DDutNIBBgyZLFFBfv3sj3l0d5eXntdGlpKUVFRQQCAWbPnsWqVSvT+r3l57dizz078v7773HiiSfjui7ffbeArl33p0ePA/nnP9/npJNO4b333mm0f69eB/POO29x2GG9WbJkMatXr6JTp3349tv5aa1TvEthTUQkgXB+K/yD0nMLB8eBUN62rwTt0uVnDBx4KYMHX47P52f//Q3Dh9/eZPs99mjPiSeezEUXnceee3aga1eTllrT6eyzT6esrIxQqIZ///ufPPDAo3Tu3KV2ueM43HXXWB566H4qKyvJycnh+utv4oUXpvLBB+/VOTcvGAwyatTdPPjgfVRVVZGdnc1DDz3G6af/mlWrVnLppefjui6FhUXcc8+4BrW0aVPIQQf14sILz+XII3/O+ecP5Oabr+eii87jgAO6s88++6b9+x8xYjT33z+GZ5+dTDgc4qSTTqFr1/255pobGDXqNqZOnUKfPkfVHtKN95vfnMO4cWO46KLz8Pv9DB9+e51RPmn5HNd1m7uGtKipCbslJeUJ2xVs+JHXrh6dsJ3P5xCJJN43Zz9yG5uLGv7llmmFhXkk8/02F9WXGtWXmlTrW7VqMe3b75PGiraKHqqKJG7YTFRfara3vsrKSrKzs3Ech/fee5v33nubMWMe8Ex9P7WWWl9jv1OKiws+Aw5Ppr9G1kRERJqJtd/wwANjAZdWrQoYOnREc5ckHqSwJiIisoOGDv0TK1duvaLWceCKK65ucJFDU3r1OoRnn30xU+VJC5HRsGaM6QeMB/zAU9baMfWWXwH8EQgDpcDl1tp5xph9gW8AG2v6ibX2ikzWKiIisr3uuef+OtNeP4wnO6eMhTVjjB+YAPQFlgEzjTHTrbXz4pq9YK2dGGt/BvAA0C+27Htr7Y5fziQikgLXdZu8KauISLLScW1AJh/kfgTwnbV2obW2GngJODO+gbV2U9xkPltuGiQi0owCgSzKyjal5ZesiOy6XNelrGwTgUBqV+9m8jBoRyD+QXC7akxKAAAgAElEQVTLgD71Gxlj/ggMAbKAE+MWdTbGfA5sAm611v47g7WKiNQqKipmw4Y1lJaWpH3djuN4OgSqvtSovtS0xPoCgSyKiopT2m6zX2BgrZ0ATDDGDABuBQYCK4FO1tp1xpjDgDeMMT3qjcTV4fc7FBbmJdxeuCR6W45kJNXOIantppvf72uW7SZL9aVG9aUmHfW1bds6TdXU5fVzmlRfalRfalRf4zIZ1pYDe8dN7xWb15SXgMcBrLVVQFXs/WfGmO+B/YFZTXUOh92k7qtU4JLU/dOSvc8aLs1yv6mWfp+rTFN9qVF9O87LtYHqS5XqS82uVF9xcUHSbTN5ztpMoKsxprMxJgvoD0yPb2CM6Ro3eRqwIDa/OHaBAsaYLkBXYGEGaxURERHxpIyNrFlrQ8aYwcDbRG/dMcVaO9cYMwqYZa2dDgw2xpwM1AAbiB4CBTgWGGWMqQEiwBXW2vWZqlVERETEqzJ6zpq1dgYwo968EXHvr22i3+vA65msTURERGRnkMnDoCIiIiKSIoU1EREREQ9TWBMRERHxMIU1EREREQ9TWBMRERHxMIU1EREREQ9TWBMRERHxMIU1EREREQ9TWBMRERHxMIU1EREREQ9TWBMRERHxMIU1EREREQ9TWBMRERHxMIU1EREREQ9TWBMRERHxMIU1EREREQ9TWBMRERHxMIU1EREREQ9TWBMRERHxMIU1EREREQ9TWBMRERHxMIU1EREREQ8LZHLlxph+wHjADzxlrR1Tb/kVwB+BMFAKXG6tnRdbNhQYFFt2jbX27UzWKiIiIuJFGRtZM8b4gQnAqUB34HfGmO71mr1grT3IWnswMBZ4INa3O9Af6AH0Ax6LrU9ERERkl5LJw6BHAN9Zaxdaa6uBl4Az4xtYazfFTeYDbuz9mcBL1toqa+0PwHex9YmIiIjsUjJ5GLQjsDRuehnQp34jY8wfgSFAFnBiXN9P6vXtuK2N+f0OhYV5CYsKl4DP5yRsB0m2c0hqu+nm9/uaZbvJUn2pUX2p8XJ9Xq4NVF+qVF9qVF/jMnrOWjKstROACcaYAcCtwMAdWU847FJSUp6wXYELkYibsJ3P5yTVDpektptuhYV5zbLdZKm+1Ki+1Hi5Pi/XBqovVaovNbtSfcXFBUm3zeRh0OXA3nHTe8XmNeUl4Nc72FdERESkRcpkWJsJdDXGdDbGZBG9YGB6fANjTNe4ydOABbH304H+xphsY0xnoCvwaQZrFREREfGkjB0GtdaGjDGDgbeJ3rpjirV2rjFmFDDLWjsdGGyMORmoATYQOwQaa/cKMA8IAX+01oYzVauIiIiIV2X0nDVr7QxgRr15I+LeX7uNvncBd2WuOhERERHv0xMMRERERDxMYU1ERETEwxTWRERERDxMYU1ERETEwxTWRERERDxMYU1ERETEwxTWRERERDxMYU1ERETEwxTWRERERDxMYU1ERETEwxTWRERERDxMYU1ERETEwxTWRERERDxMYU1ERETEwxTWRERERDxMYU1ERETEwxTWRERERDxMYU1ERETEwxTWRERERDxMYU1ERETEwxTWRERERDwskMmVG2P6AeMBP/CUtXZMveVDgMuAELAGuNRauzi2LAx8FWu6xFp7RiZrFREREfGijIU1Y4wfmAD0BZYBM40x06218+KafQ4cbq0tN8ZcCYwFzostq7DWHpyp+kRERER2BpkcWTsC+M5auxDAGPMScCZQG9astR/Etf8EuCCD9YiIiIjsdDIZ1joCS+OmlwF9ttF+EPBW3HSOMWYW0UOkY6y1b6S/RBERERFvy+g5a8kyxlwAHA4cFzd7H2vtcmNMF+B9Y8xX1trvm1qH3+9QWJiXcFvhEvD5nKTqSqqdQ1LbTTe/39cs202W6kuN6kuNl+vzcm2g+lKl+lKj+hqXybC2HNg7bnqv2Lw6jDEnA8OB46y1VVvmW2uXx14XGmM+BA4Bmgxr4bBLSUl5wqIKXIhE3ITtfD4nqXa4JLXddCsszGuW7SZL9aVG9aXGy/V5uTZQfalSfanZleorLi5Ium0mw9pMoKsxpjPRkNYfGBDfwBhzCDAJ6Get/TFufhFQbq2tMsa0A35O9OIDERERkV1Kxu6zZq0NAYOBt4FvgFestXONMaOMMVtuw3Ef0Ap41RjzhTFmemx+N2CWMeZL4AOi56zNQ0RERGQXk9TImjHmWuBpYDPwFNFDkrdYa9/ZVj9r7QxgRr15I+Len9xEv/8CByVTm4iIiEhLluzI2qXW2k3AKUARcCEwZttdRERERCRVyYa1LZdF/hKYZq2dGzdPRERERDIk2bD2mTHmHaJh7W1jTAEQyVxZIiIiIgLJh7VBwC1Ab2ttOZAFXJKxqkREREQESD6svWutnW2tLQGw1q4DHsxcWSIiIiICCa4GNcbkAHlAu9i9z7acp9aa6OOkRERERCSDEt264w/AdUAH4DO2hrVNwKMZrEtERERESBDWrLXjgfHGmKuttY/8RDWJiIiISExSN8W11j5ijPl/wL7xfay1UzNUl4iIiIiQ/BMMpgE/A74AwrHZLqCwJiIiIpJByT7I/XCgu7XWzWQxIiIiIlJXsrfu+Bpon8lCRERERKShZEfW2gHzjDGfAlVbZlprz8hIVSIiIiICJB/Wbs9kESIiIiLSuGSvBv1npgsRERERkYaSvRp0M9GrPyH6XNAgUGatbZ2pwkREREQk+ZG1gi3vjTEOcCZwZKaKEhEREZGoZK8GrWWtda21bwC/yEA9IiIiIhIn2cOgv42b9BG971plRioSERERkVrJXg16etz7ELCI6KFQEREREcmgZM9ZuyTThYiIiIhIQ8keBt0LeAT4eWzWv4FrrbXLMlWYiIiIiCR/GPRp4AXgnNj0BbF5fbfVyRjTDxgP+IGnrLVj6i0fAlxG9NDqGuBSa+3i2LKBwK2xpndaa59NslYRERGRFiPZq0GLrbVPW2tDsa9ngOJtdTDG+IEJwKlAd+B3xpju9Zp9Dhxure0JvAaMjfXdDRgJ9AGOAEYaY4qSrFVERESkxUh2ZG2dMeYC4MXY9O+AdQn6HAF8Z61dCGCMeYnoRQnztjSw1n4Q1/4ToiN2EL0tyLvW2vWxvu8C/eK2LyIiIrJLSHZk7VLgXGAVsBI4G7g4QZ+OwNK46WWxeU0ZBLy1g31FREREWqRkR9ZGAQOttRug9jDl/URDXMpio3aHA8ft6Dr8fofCwryE7cIl4PM5Sa0zqXYOSW033fx+X7NsN1mqLzWqLzVers/LtYHqS5XqS43qa1yyYa3nlqAGYK1db4w5JEGf5cDecdN7xebVYYw5GRgOHGetrYrre3y9vh9ua2PhsEtJSXmCkqDAhUjETdjO53OSaodLUttNt8LCvGbZbrJUX2pUX2q8XJ+XawPVlyrVl5pdqb7i4oLEjWKSPQzqiz/BPzaylijozQS6GmM6G2OygP7A9PgGscA3CTjDWvtj3KK3gVOMMUWx7Z4SmyciIiKyS0l2ZG0c8LEx5tXY9DnAXdvqYK0NGWMGEw1ZfmCKtXauMWYUMMtaOx24D2gFvGqMAVhirT0jNnI3mmjgAxi15WIDERERkV1Jsk8wmGqMmQWcGJv1W2vtvG31ifWbAcyoN29E3PuTt9F3CjAlmfpEREREWqpkR9aIhbOEAU1ERERE0ifZc9ZEREREpBkorImIiIh4mMKaiIiIiIcprImIiIh4mMKaiIiIiIcprImIiIh4mMKaiIiIiIcprImIiIh4mMKaiIiIiIcprImIiIh4mMKaiIiIiIcprImIiIh4mMKaiIiIiIcprImIiIh4mMKaiIiIiIcprImIiIh4mMKaiIiIiIcprImIiIh4mMKaiIiIiIcprImIiIh4mMKaiIiIiIcFMrlyY0w/YDzgB56y1o6pt/xY4CGgJ9DfWvta3LIw8FVscom19oxM1ioiIiLiRRkLa8YYPzAB6AssA2YaY6Zba+fFNVsCXAz8qZFVVFhrD85UfSIiIiI7g0yOrB0BfGetXQhgjHkJOBOoDWvW2kWxZZEM1iEiIiKy08pkWOsILI2bXgb02Y7+OcaYWUAIGGOtfSOdxYmIiIjsDDJ6zlqK9rHWLjfGdAHeN8Z8Za39vqnGfr9DYWFewpWGS8Dnc5IqIKl2DkltN938fl+zbDdZqi81qi81Xq7Py7WB6kuV6kuN6mtcJsPacmDvuOm9YvOSYq1dHntdaIz5EDgEaDKshcMuJSXlCddb4EIk4iZs5/M5SbXDJantplthYV6zbDdZqi81qi81Xq7Py7WB6kuV6kvNrlRfcXFB0m0zeeuOmUBXY0xnY0wW0B+YnkxHY0yRMSY79r4d8HPiznUTERER2VVkLKxZa0PAYOBt4BvgFWvtXGPMKGPMGQDGmN7GmGXAOcAkY8zcWPduwCxjzJfAB0TPWVNYExERkV1ORs9Zs9bOAGbUmzci7v1MoodH6/f7L3BQJmsTERER2RnoCQYiIiIiHqawJiIiIuJhXr51xy7Jlx0kmTsEb6oKQXYw8fqASFVNynWJiIhI81BY85gIcNHk/yVs5/f7CIcTx7qpg7bnPsQiIiLiNToMKiIiIuJhCmsiIiIiHqawJiIiIuJhCmsiIiIiHqawJiIiIuJhCmsiIiIiHqZbd3hMXukmhuyfnbCd4zi4rpvU+sqDuekoTURERJqBwprH+Es3s/rhiQnb+XwOkUjisOZ/5DYoUlgTERHZWekwqIiIiIiHKayJiIiIeJjCmoiIiIiHKayJiIiIeJjCmoiIiIiHKayJiIiIeJjCmoiIiIiHKayJiIiIeJjCmoiIiIiHKayJiIiIeFhGHzdljOkHjAf8wFPW2jH1lh8LPAT0BPpba1+LWzYQuDU2eae19tlM1ioiIiLiRRkbWTPG+IEJwKlAd+B3xpju9ZotAS4GXqjXdzdgJNAHOAIYaYwpylStIiIiIl6VycOgRwDfWWsXWmurgZeAM+MbWGsXWWvnAJF6fX8BvGutXW+t3QC8C/TLYK0iIiIinpTJsNYRWBo3vSw2L9N9RURERFqMjJ6z9lPy+x0KC/MStguXgM/nJLXOpNo5JLXdZHm9vmT5/b5m2W6yVF9qVN+O83JtoPpSpfpSo/oal8mwthzYO256r9i8ZPseX6/vh9vqEA67lJSUJ1xxgQuRiJuwnc/nJNUOl6S2myyv15eswsK8ZtluslRfalTfjvNybaD6UqX6UrMr1VdcXJB020yGtZlAV2NMZ6Lhqz8wIMm+bwN3x11UcAowNP0lyvbKqyzFX1aasJ1/s5+C6nDCduH8VpTntEpHaSIiIi1SxsKatTZkjBlMNHj5gSnW2rnGmFHALGvtdGNMb+AvQBFwujHmDmttD2vtemPMaKKBD2CUtXZ9pmqV5PnLSglPnpKwnZsTJFxZk3h9gy4FhTUREZEmZfScNWvtDGBGvXkj4t7PJHqIs7G+U4DEqUBERESkBWsxFxjIT8Tng4A/YbOqcCSpdvj0EA0REZFtUViT7faX2csStkn2AoizL0lHRSIiIi2XhjVEREREPExhTURERMTDFNZEREREPExhTURERMTDFNZEREREPExXg0qLoicsiIhIS6OwJi2KnrAgIiItjQ6DioiIiHiYRtakZdETFkREpIVRWJMWx8tPWPBlB4kk0W5TVQiyg4nXB0SqEh/OFRGRnZfCmshPKAJcNPl/Cdv5/T7C4cSxbuqgPmmoSkREvEzHeEREREQ8TGFNRERExMN0GFTkJ5RXuokh+2cnbOc4Dq6b+Jy6vNJNlAdz01GaiIh4lMKayE/IX7qZ1Q9PTNgu2Qsg/I/cBkUKayIiLZnCmojU0hMgRES8R2FNRGrpCRAiIt6jCwxEREREPExhTURERMTDFNZEREREPCyj56wZY/oB4wE/8JS1dky95dnAVOAwYB1wnrV2kTFmX+AbwMaafmKtvSKTtYqIiIh4UcbCmjHGD0wA+gLLgJnGmOnW2nlxzQYBG6y1+xlj+gP3AufFln1vrT04U/WJiIiI7AwyeRj0COA7a+1Ca2018BJwZr02ZwLPxt6/BpxkjHEyWJOIiIjITiWTh0E7AkvjppcB9Z86XdvGWhsyxmwE2saWdTbGfA5sAm611v47g7WKCIDPBwF/wmZV4UhS7fDptFgRkVR59T5rK4FO1tp1xpjDgDeMMT2stZua6uD3OxQW5iVccbgkenf4ZCTVziGp7SZL9aVG9aUmXAL/98XytK3vt2muL1l+v69ZtpsML9cGqi9Vqi81qq9xmQxry4G946b3is1rrM0yY0wAaAOss9a6QBWAtfYzY8z3wP7ArKY2Fg67lJSUJyyqwCWpx/gk+7gfXJLabrJUX2pUX2q8Xl+yCgvzmmW7yfBybaD6UqX6UrMr1VdcXJB020yGtZlAV2NMZ6KhrD8woF6b6cBA4GPgbOB9a61rjCkG1ltrw8aYLkBXYGEGaxUREZFm5FvogA/8VT4IurhBIAvcoAtBaqd3xZuOZSysxc5BGwy8TfTWHVOstXONMaOAWdba6cBkYJox5jtgPdFAB3AsMMoYUwNEgCusteszVauIiIg0gwhkvesn97Essj6ORpLdyN9mF9cfH95c3ACxUFd3ujbwBcFtMO1CYMv8+tNu9DUQW1+sT6SdC6dlaD8kkNFz1qy1M4AZ9eaNiHtfCZzTSL/XgdczWZuI7Hx82UEiSbTbVBWC7OC21wVEqhI/31REMqACcl4NkjsxSOA7P+G9IpSOrCSnR5DyjdVQDU4NUONADTjV0ffReeBUOxDaMr/+tBN93TJdHu3n1DhQHbeOGqLTIaLbCyc+T7jm6zDsnrG90iSvXmAgItJABLho8v8StvP7fYTD2451UwfVvzhdRDLNWeuQ+3SQ3KeD+Nb6qOkVZtOkCqpOD0EAsguDVJWEmqe4MNEQtyW8xYLilvdurkvr/XOh5KcvTWFNREREMsr/vUPu41nkvBLEqXSoOiVExVWV1BwVBq/cXdUf/dpy6ZRL/EVUSVxQlUEKayIiIpJ+LgT/5yf3sSBZbwcgCyrPqaHiihrC+ydzQoNsobAmIjuNvNJNDNk/O2E7x3Fw3W3/JZxXuonyYG66ShORLUKQ/bcAuY9nEZztJ7JbhPLrq6m4tAZ39+YdodpZKayJyE7DX7qZ1Q9PTNgumfvA+R+5DYoU1kTSphRyXwySOykL/xIfoc4RNt9bSeV5NeDd+9zuFBTWRETSIJ1XqoKuVpWdh2+VQ+5TQXKezcK30aHmiBClo6qo/kUoeh6YpExhTUQkDdJ5pSroalXxPv88H3mPZ5H95wCEofq0EOVXVhM6XOejpZvCmojILkAjf5IWLgT/6SfvsSyyPgzg5rlUDKyh4vJqIvvqfLRMUVgTEdkFaORPUlIN2X8JkPd4FoF5fsK7RygbVkXFwGrcouYuruVTWBMRkWankT9vcjZCzrNZ5D4VxL/KR+iAMJvGV1D12xAkvjBb0kRhTUQS8kUC5Fa1IbeqkJzqAmqyKijNWkdFzgYivnBzl+cJ6bytyJb1pfPWIl6vTyN/3uJb4pD7ZBY5zwXxlTlUHxNi80OV1JzgoZvY7kIU1kR2ZVXgW+3gW+XgW+3D/0NrDl5wFrlVhXW+cmoKGu3uEqEyaxNlOespy1lHec56ynPWUxZ7ZbUfWrNLXBGWztuKQPpvLeL1+sQbAp/7yH08i+zpAfBB1a+jFw2ED9JFA81JYa0J+eXFuBGXGn8lIX8VYX+1/ppIlgtOxIeTxC98QrGvWL8d2VYd1eCLJJEMHHcHN7iTqIiFsNU+fKsd/KudWCjzxebHlm2o/0OdS3fnVCqyNlKRXUJp7hrWFC6gIruEiuwSyrNLqMzaRE44n9zytuRX7kZe1W7kVe5GYVlHOqw7iGA4Z+vqjoXsgEukvUukY4RwR5dIhy2vsXkdXNy2rv597eK8PvLXog/TRiDrXT+5j2WR9XGASIFLxZU1VPy+mkiHFvx7cieisNaI3dfvzymzhtWZ5xIh5K8m5K8iFAtwNYEqgpfuSes2Qdx8cPPcuFc3qXnkEf1X+1OKAOXgVDg4jbyyZbo87jU2L7CxmGO+vAp/OItAOJtA7NUf2TrtD2fjS/ab6gE5iVtthwIGMDmpliFfVe1nGq79bKsI+ba+D4xsR35RdvQzyyX6Gea5kBv7HPPAza33msHP1R/Oih2OLCK3qg3+Z9qQvzkrGrxWOfh+jAWyjQ2Tjxt0iewR/Qp3iVBzVDg63T4SnbeHS17OGl4beUcszDatydEXF7JCeeRV7kZ+ZVuOPuMCakpa41vuw7fCIfiZH9+bAZzquvW5OS7hWHiLdHAJx163hLnIXhHcxgf3pIXw+shfizxMWwE5rwbJnRgk8J2f8F4RSkdVUnl+jf69eYzCWiN+LFrAh4c9SLAqn2AoJxZCsgmEo++D4WwCoRwC4SycMh/+9T6csviAs31DBNH/9KMBzsnei1+sHk5NoGpreIgLiNGQWAm+CL5QMFpXZGtIqhOgwllk/XYviqoDscAVq69y+4cw3JxoWHFyHApL96oNNNXBcspzNsSFnmgAcn0hIkmMXPU47Xiqc1ttnbEjoytxfbIqSvn6rx8k7BJwA/hqA+fWfRYIZ5Nd04r8yrbRffhOK3Irfdv9mUJsn20JebFg5w924IQfhjQMiPGh0VdNViSXnMo25FUVklNVSF7scGRWqN5twOdAICsuhO0XoeboaAgLt48Q2T22rL2LW+QmDpAbwgmD2jY5UB0spzpYTknBMsIDzqCsqF4cj4Cz1sG/wsG33Id/uVMb5vzLfQT/7SN7VQAnUnefRwpcaJ/NCeuH1Dvcuq72sGvY75GRCmlxvD7ytz2ctQ65TwfJfTqIb62Pmp5hNk2soOr0ECQeFJRmoLDWGMdlRfGcpP56O/uR29hctHvdmWFwKoAyB6eMWJCLvS+Pnxd7jXsf3FhDaG0VwVAOuVVtouFwS2CMNP6LIn7UL1w7KhR9dduFCLXxbR0Jyo0Lh7GRogbL6r2SS+05RwUbfuSvV49OuF+S/ev3gKsOprwoK2G7ZPk3lDB37psJ2yVbX+3nGwEqqRt640cjy+uNRlY4dd5vGcn0bSQWBnerDYfRUcksnHpJKuyroTx7AxXZGylptZyVbedGD0dmlVCeE33t++BlbO7cbuc6hOgDd3eX0O4uHNzECEQodhh3eTTA+ZY7+Fb4yF5UTc7qAnbbvA+51W0adKsMbqI8p4Swv5qwU0PEFyLiCxH2hYg4IcK+rfMCY3cjr3UWZIGbBWS5sVdws1zcbCAIbrYbfc0Csl3cYOw1ri1ZQDXRI+vp/CwiQA04NdFXapza906NU2dZ/enattXghBz8Ja0xi/vic/34IoGmX/ERIYzrRIg4EVwngus0nPY/Xkhuqyzwu+AHNxD9bAkQnY7NJwCuPzoPvxudrtMuOt8pz2a3TfsQiW2r7vai24wQweeHyJYfm9g/Yad2p299x2o/vion2sat277OaxLLHCCwsZLye/4StxWn0Rp8jkNkS1hz6/4wxFVHYOgf8Bfk1/286/9Kqjfd4EcrbrkTDNAgI9afroLwK9D2hXycSoeqfmEqBldTc3QEHIf4pOapw7SisJYRfnBbAa3c2L+V5EcqCjb8yD+uvr/RZY7rbB09I0C1Uxn9T8lX0+R/EI2GSdl+PiB2iHN7P9N4BRt+5O+NhV2X2tAWCGcTyqqkyleW+D/+osjOFdSSFYBIR5dIR5dQ3JlCvg0/8lZs//nCQfKqCsmvbFt72DWvcjfyqgrxR7LwRfwEQ3n4I358kSD+SACfG8AXCeCPBPE/n0NetdNgBG/HFXABzxB2aqLhsDYo1hBxwkRqg2IEX8SH4wbwRfxNhiZ/9yA54XR+uDn05vw6cyJOmIgTIuLb+uo6Lo7r4Iv4cfDhuNEvn+uPzndj/208lO5BmHx+yR3pW92x6T7FohVncm/6VndG89z5ojoQ4e+HreIvxy9j6R7l8C3Rr3o8cZhWaims7URcxyUUqCQUqEx6ZEh2Eg7R4O2vpopSfD6HpM5m3oVF/DWU5q2hNG9Ng2XJ/Puo/UMmDFSBUw1UO7FXcOq9pyo2alUVN2JVFRvNqoLsTWV883//jgZDNy4gbvlyA7WBMRIIxQJcw7C05dX0O4qqgrzoCF4gNroXJPYaPx0b5QsQHekLuLFXoqOFsfn5ZWuZfut9cdtp/JB3Ur9bXIezH7iVza13xwkT3YchJ3pUIUL0oqEwsWVOvenocieydT5hyNu4kf8+/kpcOPThuP6499HpOiNX0WJiJdWdd2j/06jMj514FTcQtvV1Sz8a/sET3zb2Pqd8E59OfYP4obfaLW5ZFy6OD9w6/3br7sstg21HDTqbivw2TW876eno+rPD1cxfuDq+qobtHdjctYxAYQ3nRHvRlOY8TCsNKayJyK7NT2zUFOr8B7ydo6eBDSXMmftGwnbJ/qH1s+sN5UVp/BW9IUx1sCw963Jih35zabC/Gn5nye3HyIZylr3+ecJ2ye6/Xv2PobL++ZIpyNpQyqJ3P07YLtn6+vziVKqL8hO2S1b2hnV8/eI9Cdv5/qdbs+yMfurrEEVERERkOyisiYiIiHhYRg+DGmP6AeOJHmh4ylo7pt7ybGAqcBiwDjjPWrsotmwoMIjoGQ3XWGvfzmStIiIiIl6UsbBmjPEDE4C+wDJgpjFmurX2/7d37/GWznX/x197hsZxnH6UpJL0luQ0KpW6K+lwk1MOQ6RwV/cDEZJKOZS6S6nQnegg3TdSIlSTHFPq1mAcordISiemJCaUMb8/vt+197XXXmvvwezr+10zn+fjMQ9rXdfasz9m73Wtz/U9fD63NF62L3Cf7edKmgl8AthN0gbATOAFwNOBSyQ9z3Y0IQwhhBAGzHIPP8jUeQ9O+LqpD0xlxX9O/FE/f/kV+McyK0z4usXFZI6svRi43RoDpnoAACAASURBVPavASSdDWwPNJO17YGj8+NvASdLGsrHz7b9CHCnpNvz3zfx6s4QQgghVGXqvAeZ/+WvTPi6BcsszfyHJ67vNnXffSCStUViLeB3jed3A92FW4ZfY/tRSfcDq+XjP+v62rUmL9QQQgghTJopU2Cpifs2PzL/sYV6HVMW7ZL7RTnyNxmjfkML0xbjiZC0M/AG2/vl53sBL7F9QOM1N+fX3J2f30FK6I4Gfmb7f/LxLwPft/2tcb7lvcBdk/H/EkIIIYSwiD0LWH1hXjiZI2u/B9ZuPH9GPtbrNXdLWgpYibTRYGG+tttC/Q+HEEIIIQySySzd8XNgPUnrSHoKacPABV2vuQDYOz/eGbjM9oJ8fKakaZLWAdYDrpnEWEMIIYQQqjRpyZrtR4EDgB8AtwLn2P6FpGMlbZdf9mVgtbyB4BDgiPy1vwDOIW1GmAXsHztBQwghhLAkmrQ1ayGEEEII4cmLDgYhhBBCCBWLZC2EEEIIoWKRrIUQQgghVCyStQEjaTVJO0qaUTqWEGol6amlYwghhEUlkrVxSJoi6S2FY7hI0ob58ZrAzcA+wNclHVwytg5Je/c5vrSks9qO5/GQ1N1VoyqSli4dQ1PNNwuSVpa0r6RLgetLxzMIJK063p8K4hvka0vcMIRFZjKL4g4MSdOB/UktrS4AfkgqO3IocAPwv+WiYx3bN+fHbwd+aPutklYEfgJ8tlxoww6SNM32qZ0DkpYHzmN0y7EafRN4ZukgmnJ/3NcAewDbAsUu+pIuAo6wfXO+WbgOmA2sK+lU20V//yQtS+olvAewKbAisAPwo5JxTUTS1sDhtrcuHMpcUju/R/Pzoca5BcBzWo9otIG6tkhaGXgz6ffx+cDTC8fzWdsH58cH2f5c49zptt9WMLabSL9jHQtIv4+XA5+y/XCRwBoknUSfGG3/uM1YIllLvg7cR2oUvx/wAdJFawfbc0oGBjQ72m4FnAZg+wFJj5UJaYzXArMkLWP7REmrA98DLrV9ROHYJjI08UvaIWkL0kV+B2BV0g3EYUWDqvhmQdKZwCuAi4GTgMuA221fUSqmbpJeA5xC+tA+H/gE8FXS791xBUPrOBF4NelneRbw41yYvBbVX1sqv2F4ZePx3sDnGs83ajmWbtv2OLYqKc6TgP9oN5yeZvc4tipwvKRvtHmzGsla8hzbLwSQ9CXgj8Aza8jsgd9JOpB097sZqUhw5wJRxRSZ7b9Kei3wfUlPJ124TmnexVWs+AeTpI8BuwC/JX1gHgPMtv21ooElNd8sbEC6yboVuNX2fEnFf55dPg28g3Qj+Mb83yNsn1w0qsz2wXkk91XAXsBJki4GvmD7zqLBUf+1ZQBuGIb6PC7Odq9e3ncB10uqYhlDv2uwpFOAq2nxZjWStWT4Aylf8O+uJFED2Bc4lnSHuZvtv+XjW5Du0IuTtFN+eCpwAnApKcncCcD2t0vFBiDpQnonZUPAai2H08t+wG3AF4ALbT9SUdJR7c2C7U0krQ/sDlwiaS6woqSn2v5zydgaFjQ+uM+X9PtaErWOPJJ2ef6AnAl8BPgVOTEvqfZrC/XfMEyRtAppfXrncSdpm1ourAlVvZ7e9kOSWv2ekawlG0v6e348BCybnw+RLrbTSwVm+x7gXT2OXy7pqgIh9fKmxuMLuo4tAEpfUD/1BM+1ZU1ga1LS8VlJl5N+B5fKbdtKqvpmwfYvgaOAoyRtTvo3/Hm+4XpZ2egAWLmRcAAs1XxeOtnI67+2B3YDVie9V2fY/m3JuBqqvrYMwA3DSsC1jCRo1zXOFU0qJW3W4/AqwJ7UMYXck6SlSKPQd7f5faPdVOUk/dj2lvnx123v1Th3ne1ev/ChQdJ023/vc+6ZFX0wIWkaaS3H7qTplUtt71E2qsGSp/VeYbv4BV/SeAntAtv7tBZMD5LmkUbRzs7/HfWBUDqZHDSNG4ZdgFpuGKqUb0qbFgB/Aa4ATrX9rzFf1DJJD5Diak4h/wO4EjjY9h/aiiVG1vrId5w7Arvb3qZgKMs3Hr+g61wVaxAkHQLcb/vLXcf3BVYsvWOQ9ObfDEDSpba3apw7v3OuBrYfAc4Fzs2L+HcsGc84U8gA2N6uxXDGkPRq4ECgMydxK3ByLWuGbL+9dAwT+Cbp5ytG/g07io9cDcC1ZRTbs4HZkg4j3WwVJWncne4lb1Rtv7rU915YtlcsHUNHJGsNkp4CbEPa1fN60ofmKUWDGn+oupZh0beQpsW6fZ20m6b0BbWZ1HbXjiqe8Ep6BvDszlbw/AG1Qj59ZrHAkhqmiXuStA1wMmma9hjSz3Iz4CuSDrD9vZLxdSjVSXwvIzdbvyCVJripXFTJeKUbKqkTVvu1pfYbhu8ydmRoAWnKew0Kr1ur+b3RkfOCtzA6xjPzjXVrIlkDJL2ONHT9OlKNlzOAF1VyV7yypB1JCy6b61+GSOsRarBUryFr2//MU1KlddfJ6XeulOMZXcvvnaQF1cuRkpBihZltX1nqey+E95LK69zQODZH0mzSzrziyZqk7UkJ78dJO0MBNge+Lekw298pFlwPtdUJo/JrS+03DJ0qBx2Sng28j7QG9WMlYmrEUv17Q9IGpLWSPyGt/YO0c/qDkrazfUtbsUSylswCrgK27GxXl1TF1nDS3Ph2jcfNBbfF1+RkU3otqK3kzhxgjTxaNdR4TH6+ermwhsn2RY3n/7D9aYDSm0gqL1z5tK5EDQDbN1b0u3cssLXt3zSO3SjpMuA7+U9RldcJq/3aUv0NA4Ck9YAPAi8hJUbvrmBNWPXvDdLP8D9t/7B5MJeT+TypRmErIllLNiNtWb9E0q9Ji22r2NZcyejeRI4HvivpUEZ2G83Ix2uYRjuN9AHU/RjgS+2HM8YyXc+ba+r+X5uB9FBz4cp5T/Bcm5bq+jACwPZvVEErsQGoE1b7taXqG4Y8zfhB0hTeJ4F9bc8vG9Wwqt8b2VrdiRqA7UuUuhu0JpI1IHcpmAMcIellpCnRpSV9Hziv2eqkbY1RoJ5sn9BWLOPEcIake0l3ShuSRl9+AXzY9veLBgfYPqZ0DBN4QNLzbN8GqRAoQC4J8EDJwCovXLmupAt6HB+ifJukjkd77TiW9CxGWjyVVHWdsB7XFkj9kau4tlD/DcMNpLZc3wVeDLy4WR/M9rsLxQX1vzcgjexO616fJmkZWs6fIlnrYvtq4GpJB5FrS5HWD5VSzW6U8eQL55iLp6QX2f55gZCaMSxD+jneB1xImrp4JXAH8BHbcwuGB6lO2EWSjmP06MEHgIOKRTWx0oUrtx/nXA2jLpB+tpcodanorHnZHDiCtHaoqAGoE9b32lKJ2m8Y9qWOdbm9VP3eyM4g7czfv3Pjmtf9nUja5NKaqLNGajFle78ex9cGZtnuLpkRxpEXZe6e//zN9uaF4zmH1KVieVLRxZtJSduWwCa2e031tSpPVxzOyI6jm4HjPdKXs4gJClc+aPvAlkMaOJI2Bg5l5Gd7C2m935jps9IkzSC9b3elgjphGttIe5TCI0NI+rfxzle+Qae4QXhvSDqAdG1ejpSEP0iKsdVp0EjWAEmnk0YZ32r7sXxsA9IH+rEu2KNR0onjnS99serIdxudBO1fwLOAzXutSWibpJttb5grT99t+2mNczfY3rhgeFWruXBljq3fBWxBVz290EPesTim/VUthYUl7T3e+ZLX5okoNfrerXAMVddJHCS57iW2iyxNiWnQ5O3AF4FvSJpJ2jHzDdIukIvG/crJd23j8TGkoeOqSPopMJ20MePNtn8l6c4aErXsnwC2H5XUXXG6+GLbHhfU4R2Xtv+nTFRJ5YUrD+txbAvSXfA9LcfS0wB8WO5DKj0xilO/0Bp2g8r2B0oH8QS9tHQA1LMcYIwBeG+MqoFp+wFJh0garoFp+/a2YolkjeEL0zvyKNYVpFGhXWz/rGhgjL5zlHRwpXeSfwbWAp5KKoUxpm1NYc/IP9uhxmPy87XKhTWs1wV1VWBPSRvaPqLtgJpqLVxpe/hGJk9HfYi0s/ZdlSw+h5Gf7RBpJ/KY5RZhXG8grd0MT8B407CSXt5mLD1Um0g2VFMDM6ZBGbUuYohUa+g60u4ooKqpxmp7gUpaCdiJNA26HrAy8Hrb1xQNjMGdSpE0FbjW9iYFY2gWrpydD28OvB8oXrhS0uuBI4FHgONsd0/bVkPS9bY3LR1Hk6RHSb0Ouw2RppKntxzSKJJuIBUh7VkAt7NzupQ+azohxXuR7TXbjKdbvobsSropnWX7ZknbkhLgZWv7faxN92du8z0s6SrbrbUUi5G1ZHafx2Eh2b4f+CrwVUlrkC4Qn8lbs9cuHFvfZEwT9M4rKZdRKB1GtYUrJf2cNJJ7PPDTfGz4wmr7uj5fWkqNd8Y3Vf6BvT5pKUivZG0B5Xdcfnqcc79sLYr+vgysDVwDnJiXgWwOHGH7/JKB9Si4PYrtjVoMp59qamBGskb/D/Nc8uFNvc61RdI8RtZVLSfp7/lxFXe+vdi+h7QO5mRJ55aOB0DSS0l3lz+yfY+kjUhbxF9BupiVjK27XymkHZdvJU05llRz4cp5pJ1ZO+c/TQuA17QeUZeun+1USavQSDxKjwwNgFtqTiYrX9MJKTHbyPZj+fPsT8C6tv9SOC4YKbg9RKoD9+8FY+mnmhqYkax1ycPGr2ekV+hVwDcLhnRbzRerhfCi0gFIOp50YZgDvE/SD0hrhz5OWmBd2rWMbrbc2XF5OfCfpYLKqi1caftVJb//Qur+2TZH+2oYGep7bauhRuIgyDMJ+zN6Tefn801raf/sVDiw/bCkX1eSqI0quC3pkT4FuEurpgZmJGtZXqC8Bym7vwZ4ObCO7V7rOdpU49TJ41G82TKwDbBpvlitQqrovWEtu1Vtr1M6hnFUW7hS0k5dhzq7aOeU2l7frfKfLbZHNfPurpFI+lmX1LNHc2fWw3bJG+nOIv0zgdNJBVQhfZhfI+kttn9SKrZsfUk35sdDpCK+NzIyM1PDVGO1bM/K15nDgc7a9ZuBndqugRnJGiDpbuC3wBdIi6YfyKUnSidqMLrx+BiuoN1Un2k8SBeEGpK1h50bjtu+T9KvaknUOmq9O7d9vqQ7SYUrOwVwfwHsWkHhyl5LFFYFNpK0r+3L2g6oW94AsaLtb3UdfzPwd/foO9i2mmsk2j6987jCWQ9Ia9Z2sN1svXaBpPNI5aBeUiasYc8v/P376tqcsaykTRm9RKCKNac5KXtr6TgiWUu+BexAakk0X9J3qGdEayqwAnUkPf10T/U0/bPlWHp5TldLmHWaz0vX86n97jwnZWMuVpI+ZbtXrbNW2H57r+N5ivYcyn9QAnyYdG3pdiWp6HbRZG0AaiTWPOsBML0rUQNSv+lOEdXC1qyhBFUfzc0ZfwKaAw+1rDmtpgZmJGuA7YMlvYe0RXx34JPASpJ2Bb5n+8GC4f3R9rEFv/+Eap/qYWwPyfF2cJVQ+915P7vSuzBtUbbvqmDzQ8c02/d2H7Q9V9LyJQLqUnWNxMpnPQCGJK1i+77mwTzbULp3LsB/A5tBSsxt11CoFxiIzRlQUQ3MSNayXBj3cuDyfKF/AzCT9Mve6hbdLjWPqA2T9BRSgcDmNN6Zth8pF1UyXmHIStR+d95Plb+bSvVOiv/eZdMlLWV71GaMfI1ZtlBMw2zv0KiReLSk9YCVJb24hhqJ1D3rAfAZ4GJJhzF6AfongM8Wi2pE8z3aXYaiOEmrkUZN18+HbiV9blSxS7rfZ0eembmWtHa3FVEUt4uk1QE6d8OSlrX9UMF4Vq3lF7efvCj5AuAnjCxCn0GartjedtHyEz3q+QwPZZMq8T9cJLBM0q3Ay/rcnV9te/3eXzn5JliPeIPtZ7QZT1OPKQpId71rAnva/mn7UY0m6b9Io1YH2J6Xj61AWjg/13bRTRrdJD2VNGI6EyheIxGG+5S+ijTr8e/ASsC+lJ/1ACAXmT2cdKO6gNSM/HjbFxYNjFFFhacAl9FVYLjkZ4uk5+eYfgBcn+PaFNgaeI3tGurU9SVpTpsFy2NkjeGLwVHAAaQ1Yp3K3ieVnoKsPVHLTiL1UR21/kbSa0n11koPd2/b49iqwN6k2P+j3XDGqPnuvOb1iN1TFJ2SJ7+yXTq2jiOBjwJ3SbqL9O+4NqlY6YdKBtaL7T+T3hMn5bV/xfWY9ehsMig964GktZ36R4/pIS1pW5fvLb0So4sK11Q65iPAQbbPaR7Mm2+OA95cJKrRsVRTAzNG1oC82/KNwDts35mPPYe0TmKW7c+UjK92kn7Zb/RH0q22a96RVEULoJrvzmsnaR1Gpt9vsf3rkvH0ImlZ4Ln56e0lR+ubujbejFHB5psxNf4a54rOeuQYfgm8oXtDhqR9gA/aXrdIYANAkm33bNEy3rk25Z3w/WpgftT23/t97aIWI2vJXqSWOnM7B2z/WtKewMWkkY/Q3xRJ07rXp+VaSLX/jtWwCJhx7s4Ptl16dG0USUfbPrqCOFYkjVDNADplRDaRdC2wb5sX0n4kvbLH4RcptxGz/aN2IxrjpaS6g2cB/0d96xDPZ2SB/Lm2h0dbSidq2SGkUfFtbP8KQNL7Seuw/q1oZCmWA2yfnB+/oPSSlC7znuC51tS0ea72D9K2LN1M1Dps31vRrrKanQGcK2n/ThXqXLvpRODrJQODvs2WVwH2BEp/WE7kEMpPhXbbDji6dBCk6bpbgJnOVdrzkoYPkabfi9dGAt7b49gCYCPSdOjUdsMZ42mkNUK7kxKM7wJnVfSh3kweS3d7GMP29yQ9Anxf0g6kzigvBl7ZvQa1kH1I7wVI1+J+jedL6FdDdIi0M7kKtdTAjGQtGW99Sy1rX6pl+6OSDgCukrRcPjyPtHj/pIKhdXSX6ugMZV8BnNp6NI9PbSMdUE9ML7f9tuaBvL7pWEm/KhPSaLZHFe7NNfWOJNWVOrDnF7XI9nxgFjBL0jRS0naFpGM6IzKFdW8Mqo7tSyW9nXQ9uZq0OL7opqU+annfdpwG9Nvt/qU2A+mnphqYkawlG2ukQXrTEBVud65RvrCf3Ck14dzuR9JT86LlkrGV3uDwZBT9gGpOozTMKBLM41PVB5OkrUgjfguAj9XQuaAjJ2nbkBK1Z5NGxM8rGVND59o8RKpy37lOd9olTS8XGkh6gJE1TdOArYB78ghv8fhIZVh2JC33mK6uFm22v10mLLB9TKnv/ThUUwMzNhiERU7SyqSdPHsAz7f99MIhIWlD0pRUcyj7U7ZvKhdV0rjgdxsClrVd7KZK0nW2a5o6GSbpa8AdwEfyiFrn+IeA59neq1hwI7FsA3wQuB84zvaPC4c0iqQzgA2B7wFnu+V+h2FySfrqOKcX2N6ntWC6SDrH9q758SeaZWwkXWz7daVia8Rxi+0NHu+5yRAja2GRyLvdticlaJuShrd3oII1YZK2J5V5+DgjU6KbA9+WdJjt7xQLDrBdc+Hbmh1I2mBwu6Q5+dgmpJpN+xWLarQLgbtJ0+6HSzq8ebL0bkvSus15wEHAuzsbH6hk5Co8Oe7Tkg2Ga+qVtF7j8dZAs+ZgLWvWqulQEclaeNIknQm8grRz9iRSocPbbV9RMq6GY0m7fX/TOHajpMuA7+Q/obeNxlkiUPTDPO/23EXSukDnDvcW23eUiqmHqqfgbVexGzq0o3vWAyg56zHetF4tU37j1cBstUpEJGthUdgAuI/UKuRW2/Ml1fJmA1iquw4SgO3fxG7fCd1UQx26XiS9HljR9rdI06Gd4zsD99ewLmwAWp2FxVzFsx7LSdqUNEK1bGPX/hAVtGIDsH2qpD+QCvg2l9B8tO0amJGshSfN9iaS1ictUL5E0lxgxRo2F2SP9iqumSu0P9rna0L9Pkz60Ol2BWn6sXiy1qPV2Si2N2oxnLCEqXzW44+kZSlDpN3RzY4kfyoSUQ/9amC2LZK1sEjkPm5HAUdJmkFK3H4u6W7bLysbHUeRksiPMdK7dHNSE96qejNW6JulAxjHNOcevk2250pavkRAPfRqdRZCW2qe9Xgf8DvbfwSQtDdpivY31FHHEQBJrya1omw2mz+57YQ3doOGJy33x/tdj+NDwJG2P1IgrO5YNgYOpdGWiLQb9Ib+XxUkHUX/kaEFJX+2km4DNrD9aNfxpUlr19br/ZUhLDkasx67AXMBARuWnvWQdB3wWtt/zZ0+ziZtGtqEVEVg55LxwfBu7pNJ656vI40CbkaqlXiA7e+1FUuMrIVF4YeSxvTHA96W/xRP1nJSVkNF+0HzYI9jy5F2W65G2Z/tt4HTci24eQCSVgA+l88V16MsywLSB+blwPts/6VIYGGJUfGsx1Tbf82PdwNOtX0uqRvOnHG+rk3vJdVZa97Uz5E0mzStHMlaGCi198e7kPHXDZUun1At28PdH3LB44NILWzOZmxniLYdCXwUuEvSXfnYM0nlPD5ULKqGXmVZJK1Cuok5Bdil7ZjCkqN71sP2tcC1kt5Lev+UNFXSUnlkfCvgHY1zteQmT+s1+2L7xrZLn9TyDxIG2AD0x/vUxC8J/eSaQocAbwG+BmxWw881X+SPkHQM8Nx8+HbbD9W8yzf/231GUvGivWGxV/Osx1nAlXlD2kPAVQCSnksqIl2DaprNR7IWFoma++M1yydIWj0fG7MwPYwl6XhgJ1IP1Rfa7jUtWpTth4Cb8hrJ10jag7Swv3TRz75yMhnX3zDZqp31sH2cpEuBNYGLG11IplBB39xsXUkX9Dg+BDynzUBig0F40nr0x/sXMJ8KCqd25IXyB5IuBEOkkh0n2T62aGCVk/QY8Ajp36t5sajpZ7sF6cNnB2BVYH/gghpG/7p7MWarkNbo/Dh+/8Jky31pv0h6f3RmPbap4f1RO0njJrRt1lGMZC0s9iQdArwReIftO/Ox5wBfAGbZbrUSdVg0cimWXYDfkqZUzgNm216naGANPXozLiC1nrrC9ncLhBSWQJJeQXp/XA3sWsusR+0kTc+dUnqdG1O7czLFMHxYEuxFajc1t3PA9q8l7UkqFhnJWh95vVpfjd1cJewH3EZKui+0/UhFNaSA8XszhjDZesx6bAXck5cMVDEyXrkrSKU6kHSp7a0a587vnGtDJGthSbB0M1HrsH1vzQvRK3EtIxf7bgtoed1GlzVJDaB3Bz4r6XJS25rODrPiJL0AWNf2Bfn5Z4CV8umTbV/X94tDeJJ67UYOj0vzutd949rrmjhpIlkLS4J/PsFzS7yaphS72Z4PzAJmSZpG2lSwLPD7fBe8R9EAk/8CPt54/npSWZHl6N8uK4RQh+4aif3OTbpI1sKSYGNJvdYdDAHLtB3MoJO0LmlB/0zbL5jo9W2w/QjQKai5IrBj4ZA61rR9deP533PhTyS9s1BMIYSFs0Ze8zzUeEx+vnqbgUSyFhZ7tqeWjmHQSXo6aQfjHsALSaNFM4sGxfBurftykcpdgVcCdwD/XTayYaOmoWxv0Xi6RsuxhBAen9MYeQ83HwN8qc1AIlkLIfQl6R2kNWFrAecA+wLfsX1M0cAASZ8HNgKm5T6hK5CmRV8OfIVUxLe0P0h6ie3/ax7M5Ub+UCimEMLC+UqvvtcAkrZtM5BI1kII4zkZ+Cmwh+3ZABXtuHy17Q0kLQP8HljD9nxJXwRuLBxbx/uAb0g6ndQIGmAGsDdppDKEUK+eHSByAfgjgYvaCmRKW98ohDCQ1iTVMPu0JEv6CFDLDtqHAXLNqLvyhgNyJfR/lQysw/Y1wBbAVEZa/EwBtsjnQgj16nSAWK9zIHeAOISWO0BEUdwQwkKR9AzSaNDuwPLAebY/UDCeu4ETSIt935Mfk58fbHvtUrGFEBYPtXSAiGnQEEJfkl4E/M72n2zfnZsu/4E0UvRQ2ejqWfzbj6Sb6L3Fv1OUdKOWQwohPA619L2OZC2EMJ4vAq8FkPRK0i7QA4FNgI0LxkUNmxwWwudJmx56JWyxZi2EitXUASKStRDCeKY2WkrtBpya64SdK2lOwbiQdOJ4522/u61YxvF54EpgL9u/b56QNBP4ZJGoQggTqqkDRGwwCCGMZ6qkzk3dVsBljXOlb/beBWxJmpadTWqN1fxTgxtJGzR+JmnnrnOttqsJIQyu0hfbEELdzgKuzGvVHgKuApD0XOD+koGRdqruQhrxexT4BvAt238rGtVoC2yfJulK4H8lbQPsb/sftNyuJoQwuGJkLYTQl+3jgEOB04Etc1kMSNeOA0vFBWD7L7ZPsf1q4O3AysAtkvYqGVcvtm8DXgr8Gbhe0ksKhxRCGCAxshZCGJftn/U4dluJWHqRtBmpnMjWwPepZwoUGlOdth8FjpA0izRi2WpvwRDC4IpkLYQwkCQdC2wD3AqcDbw/J0Q1GbNj1fYVkmYA0cg9hLBQoihuCGEgSXoMuBP4Rz7UuZhFDbMQwmIlRtZCCINqndIBhBBCG2JkLYSwWJG0JbC77f1LxxJCCItCjKyFEAaepE2BPUilPO4Evl02ohBCWHQiWQshDCRJzyPtAt0dmEuqszaUS3mEEMJiI5K1EMKg+iWpSO+2tm8HkPSesiGFEMKiF8laCGFQ7QTMBC7PtcvOJlo4hRAWQ9HBIIQwqC6yPRNYH7gcOBhYQ9IXJL2ubGghhLDoRLIWQhhU1wDYnmf7TNtvAp4BXA+8r2hkIYSwCEXpjhDCQJJ0ve1NS8cRQgiTLdashRAG1eqSDul30vYJbQYTQgiTJZK1EMKgmgqsQGwqCCEs5iJZCyEMqj/aPrZ0ECGEMNlig0EIYVDFiFoIYYkQyVoIYVBtVTqAEEJoQ+wGDSGEEEKoWIyshRBCCCFULJK1EEIIIYSKxW7QEMJiSdJ84CbSde5OYC/bf5P0bOBWwI2Xn2D7DEkrAMcDrwPuBxYAp9g+LX/dRbY3zH//lsAJwPTG33FqPnc0cDjwbNv35GMP2l5hEv+XQwiLqRhZq92PRgAAAedJREFUCyEsrh6yvUlOrv4K7N84d0c+1/lzRj7+JeA+YD3bmwFvAFbt/oslPQ04E3iX7fWBLYF3Stqm8bK5wKGL/n8rhLCkiWQthLAk+Cmw1ngvkLQu8GLgSNuPAdi+1/Ynerx8f+B029fl180ljaQd0XjNV4DdJI1J9kII4fGIadAQwmJN0lRSmY8vNw6vK2lO4/mBwCrADZ1EbQIvAL7WdWx2Pt7xIClhOwg46vHGHUIIHZGshRAWV8vmhGwt0hq1HzbO3WF7k+aLJW3X9fyDwC7AGraf/gRjOBGYI+lTT/DrQwghpkFDCIuth3JC9ixSt4P9J3j9LcDGkqYA2D4uf/30Pq+d0XVsBvCL5gHbfyOtbZvoe4cQQl+RrIUQFmu2/wG8GzhUUt/ZBNu3k6YyP5qnTpG0DL3bWn0eeJukTfLrVgM+AXyyx2tPAN5JzGSEEJ6gSNZCCIs929cDNwK750PrSprT+PPufHw/YDXgdkmzSVOnh/f4+/4I7AmcJumXwNXAV2xf2OO1c4HzgGmL+v8rhLBkiHZTIYQQQggVi5G1EEIIIYSKRbIWQgghhFCxSNZCCCGEECoWyVoIIYQQQsUiWQshhBBCqFgkayGEEEIIFYtkLYQQQgihYpGshRBCCCFU7P8DzISGs42G1fkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,5))\n",
    "train['REGION'].value_counts(normalize=True).plot(kind='bar', label='certain_region', alpha=0.8)\n",
    "train[train['CHURN'] == 1]['REGION'].value_counts(normalize=True).plot(kind='bar',\n",
    "                                                        color='r',alpha=0.5,\n",
    "                                                        label='churn == 1 & certain_region')\n",
    "\n",
    "plt.plot(train[train['CHURN'] == 1]['REGION'].value_counts().values / \\\n",
    "         train['REGION'].value_counts().values, color='magenta', \n",
    "         label='(churn == 1 & certain_region) / (certain_region)')\n",
    "plt.legend()\n",
    "plt.ylabel('counts')\n",
    "plt.xlabel('REGION')\n",
    "print('(churn == 1) / (all_dataset): {} %'.format(train['CHURN'].sum() / train['CHURN'].shape[0] * 100));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(churn == 1) / (all_dataset): 18.75473527052322 %\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,5))\n",
    "train['TENURE'].value_counts(normalize=True).plot(kind='bar', label='certain_tenure', alpha=0.8)\n",
    "train[train['CHURN'] == 1]['TENURE'].value_counts(normalize=True).plot(kind='bar',\n",
    "                                                        color='r',alpha=0.5,\n",
    "                                                        label='churn == 1 & certain_tenure')\n",
    "\n",
    "plt.plot(train[train['CHURN'] == 1]['TENURE'].value_counts().values / \\\n",
    "         train['TENURE'].value_counts().values, color='magenta', \n",
    "         label='(churn == 1 & certain_tenure) / (certain_tenure)')\n",
    "\n",
    "plt.legend()\n",
    "plt.ylabel('counts')\n",
    "plt.xlabel('TENURE')\n",
    "print('(churn == 1) / (all_dataset): {} %'.format(train['CHURN'].sum() / train['CHURN'].shape[0] * 100));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "'K > 24 month', 'I 18-21 month', 'H 15-18 month', 'G 12-15 month', 'J 21-24 month',\n",
    "'F 9-12 month', 'E 6-9 month', 'D 3-6 month'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(churn == 1) / (all_dataset): 18.75473527052322 %\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,5))\n",
    "\n",
    "# смотрим на те тарифы, у которых больше 2000 пользователей\n",
    "filter_array = (train['TOP_PACK'].value_counts() > 2000).values\n",
    "\n",
    "filter_array_churn = (train[train['CHURN'] == 1]['TOP_PACK'].value_counts() > 2000).values\n",
    "train['TOP_PACK'].value_counts(normalize=True)[filter_array].plot(kind='bar', \n",
    "                                                                  label='certain_pack', \n",
    "                                                                  alpha=0.8)\n",
    "train[train['CHURN'] == 1]['TOP_PACK'].value_counts(normalize=True)[filter_array_churn].plot(kind='bar',\n",
    "                                                        color='r',alpha=0.5,\n",
    "                                                        label='churn == 1 & certain_pack')\n",
    "\n",
    "churn_certain_pack = train[train['CHURN'] == 1]['TOP_PACK'].value_counts()[filter_array_churn].values\n",
    "plt.plot(churn_certain_pack / \\\n",
    "         train['TOP_PACK'].value_counts()[filter_array].values[:churn_certain_pack.shape[0]], color='magenta', \n",
    "         label='(churn == 1 & certain_pack) / (certain_pack)')\n",
    "\n",
    "plt.legend()\n",
    "plt.ylabel('counts')\n",
    "plt.xlabel('TOP_PACK')\n",
    "print('(churn == 1) / (all_dataset): {} %'.format(train['CHURN'].sum() / train['CHURN'].shape[0] * 100))\n",
    "2000 / train.shape[0] * 100;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Check summary of numerical fields\n",
    "#train.select_dtypes(include=['int64', 'float64']).describe().T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "REGION             849299\n",
       "TENURE                  0\n",
       "MONTANT            756739\n",
       "FREQUENCE_RECH     756739\n",
       "REVENUE            726048\n",
       "ARPU_SEGMENT       726048\n",
       "FREQUENCE          726048\n",
       "DATA_VOLUME       1060433\n",
       "ON_NET             786675\n",
       "ORANGE             895248\n",
       "TIGO              1290016\n",
       "ZONE1             1984327\n",
       "ZONE2             2017224\n",
       "MRG                     0\n",
       "REGULARITY              0\n",
       "TOP_PACK           902594\n",
       "FREQ_TOP_PACK      902594\n",
       "CHURN                   0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Check for missing values in training data\n",
    "#train.isnull().sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocessing\n",
    "\n",
    "### Work with categorical features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#We will drop REGION, TOP_PACK, and MRG\n",
    "train.drop(columns=['REGION', 'MRG', 'TOP_PACK'], inplace=True) #drop these columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2154048, 14)\n",
      "(2154048,)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split, cross_val_score\n",
    "\n",
    "y = train['CHURN']\n",
    "x = train.drop(columns=['CHURN'], axis=1)\n",
    "#test = test.drop(columns=['user_id'], axis=1) #you will use this for predicting and submitting the resulting\n",
    "print(x.shape)\n",
    "print(y.shape)\n",
    "#print(test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "#from sklearn.preprocessing import OneHotEncoder\n",
    "\n",
    "#tenure_ohe = OneHotEncoder(sparse=False)\n",
    "\n",
    "#tenure_cat = pd.DataFrame(tenure_ohe.fit_transform(x[['TENURE']]))\n",
    "#tenure_cat.columns = tenure_ohe.categories_[0]\n",
    "#tenure_cat.index = x.index\n",
    "#x = pd.concat([x, tenure_cat], axis=1)\n",
    "\n",
    "#x = x.drop(columns=['ZONE1', 'ZONE2'], axis=1)\n",
    "x = x.drop(columns=['TENURE'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "#from sklearn.preprocessing import LabelEncoder\n",
    "\n",
    "#Encode the TENURE column\n",
    "#encoder = LabelEncoder()\n",
    "#x[\"TENURE\"] = encoder.fit_transform(x[\"TENURE\"])\n",
    "#x[\"TENURE\"] = (x[\"TENURE\"] + 1) * 3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Spliting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1507833, 13)\n",
      "(1507833,)\n",
      "(646215, 13)\n",
      "(646215,)\n"
     ]
    }
   ],
   "source": [
    "#Split training data into train and test split\n",
    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size = 0.3,\n",
    "                                                 random_state=1,\n",
    "                                                 shuffle=True,\n",
    "                                                 stratify=y)\n",
    "print(X_train.shape)\n",
    "print(y_train.shape)\n",
    "\n",
    "print(X_test.shape)\n",
    "print(y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "whole data\n",
      "0    1750062\n",
      "1     403986\n",
      "Name: CHURN, dtype: int64\n",
      "train \n",
      "0    1225043\n",
      "1     282790\n",
      "Name: CHURN, dtype: int64\n",
      "test \n",
      "0    525019\n",
      "1    121196\n",
      "Name: CHURN, dtype: int64\n",
      "0.23084096449154373 0.23084087660596403 0.23084116955767314\n"
     ]
    }
   ],
   "source": [
    "print('whole data\\n{}'.format(train['CHURN'].value_counts()))\n",
    "print('train \\n{}'.format(y_train.value_counts()))\n",
    "print('test \\n{}'.format(y_test.value_counts()))\n",
    "\n",
    "print(train['CHURN'].value_counts()[1] / train['CHURN'].value_counts()[0],\n",
    "      y_train.value_counts()[1] / y_train.value_counts()[0], \n",
    "      y_test.value_counts()[1] / y_test.value_counts()[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train\n",
      "(1206266, 13)\n",
      "(1206266,)\n",
      "+++++++\n",
      "test\n",
      "(646215, 13)\n",
      "(646215,)\n",
      "+++++++\n",
      "validation\n",
      "(301567, 13)\n",
      "(301567,)\n"
     ]
    }
   ],
   "source": [
    "#Further split X_train and y_train into train and validation sets\n",
    "X_train,X_val,y_train,y_val = train_test_split(X_train, y_train, test_size = 0.2,\n",
    "                                               random_state=1,\n",
    "                                               shuffle=True,\n",
    "                                               stratify=y_train)\n",
    "\n",
    "print(\"train\")\n",
    "print(X_train.shape)\n",
    "print(y_train.shape)\n",
    "print(\"+\"*7)\n",
    "print(\"test\")\n",
    "print(X_test.shape)\n",
    "print(y_test.shape)\n",
    "print(\"+\"*7)\n",
    "print(\"validation\")\n",
    "print(X_val.shape)\n",
    "print(y_val.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Work with NaNs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fill NAs for train data\n",
    "# делать после разделения\n",
    "\n",
    "columns_with_Nans = ['MONTANT', 'FREQUENCE_RECH', 'REVENUE', 'ARPU_SEGMENT',\n",
    "                     'FREQUENCE', 'DATA_VOLUME', 'ON_NET', 'ORANGE', 'ZONE1', 'ZONE2', \n",
    "                     'TIGO', 'FREQ_TOP_PACK']\n",
    "\n",
    "#1. MONTANT - сумма пополнения      median\n",
    "#2. FREQUENCE_RECH - количество раз, когда клиент пополнял счет     0\n",
    "#3. REVENUE - ежемесячный доход каждого клиента    median\n",
    "#4. ARPU_SEGMENT - доход за 90 дней / 3     median\n",
    "#5. FREQUENCE - количество раз, когда клиент получал доход    0\n",
    "#6. DATA_VOLUME - количество подключений     0\n",
    "#7. ON_NET - inter expresso call    0\n",
    "#8. ORANGE - call to orange   0\n",
    "#9. TIGO - call to Tigo   0\n",
    "#10. ZONE1 - call to zones1    0\n",
    "#11. ZONE2 - call to zones2    0\n",
    "#12. REGULARITY -  количество раз, когда клиент активен в течение 90 дней     0\n",
    "#13. FREQ_TOP_PACK - количество раз, когда клиент активировал top pack пакеты     0\n",
    "\n",
    "\n",
    "for col in columns_with_Nans:\n",
    "    #if col in ['MONTANT', 'REVENUE', 'ARPU_SEGMENT']:\n",
    "    X_train[col].fillna((X_train[col].median()), inplace=True)\n",
    "    X_test[col].fillna((X_test[col].median()), inplace=True)\n",
    "    X_val[col].fillna((X_val[col].median()), inplace=True)\n",
    "    #else:\n",
    "    #    X_train[col].fillna(0, inplace=True)\n",
    "    #    X_test[col].fillna(0, inplace=True)\n",
    "    #    X_val[col].fillna(0, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Design new features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Варианты\n",
    "# 1. общее число звонков\n",
    "X_train['total_calls'] = X_train['ON_NET'] + X_train['ORANGE'] +\\\n",
    "    X_train['TIGO']\n",
    "#+ X_train['ZONE1'] + X_train['ZONE2']\n",
    "\n",
    "X_test['total_calls'] = X_test['ON_NET'] + X_test['ORANGE'] +\\\n",
    "    X_test['TIGO']\n",
    "#+ X_test['ZONE1'] + X_test['ZONE2']\n",
    "\n",
    "X_val['total_calls'] = X_val['ON_NET'] + X_val['ORANGE'] +\\\n",
    "    X_val['TIGO']\n",
    "#+ X_val['ZONE1'] + X_val['ZONE2']\n",
    "\n",
    "\n",
    "# 2. MONTANT / FREQUENCE_RECH (средняя сумма пополнения)\n",
    "X_train['MONTANT/FREQUENCE_RECH']  = X_train['MONTANT'] / (X_train['FREQUENCE_RECH'])\n",
    "X_test['MONTANT/FREQUENCE_RECH']  = X_test['MONTANT'] / (X_test['FREQUENCE_RECH'])\n",
    "X_val['MONTANT/FREQUENCE_RECH']  = X_val['MONTANT'] / (X_val['FREQUENCE_RECH'])\n",
    "\n",
    "\n",
    "# 3. ARPU_SEGMENT * 3 - MONTANT (сколько оставалось после пополнений)\n",
    "X_train['remains'] = X_train['ARPU_SEGMENT'] * 3 - X_train['MONTANT']\n",
    "X_test['remains'] = X_test['ARPU_SEGMENT'] * 3 - X_test['MONTANT']\n",
    "X_val['remains'] = X_val['ARPU_SEGMENT'] * 3 - X_val['MONTANT']\n",
    "\n",
    "\n",
    "# 4. DATA_VOLUME / REGULARITY (плотность активности, т.е. число подкючений на число входов в сеть)\n",
    "X_train['activ_density'] = X_train['DATA_VOLUME'] / (X_train['REGULARITY'])\n",
    "X_test['activ_density'] = X_test['DATA_VOLUME'] / (X_test['REGULARITY'])\n",
    "X_val['activ_density'] = X_val['DATA_VOLUME'] / (X_val['REGULARITY'])\n",
    "\n",
    "\n",
    "# 5. общее число звонков / REGULARITY\n",
    "X_train['calls_density'] = X_train['total_calls'] / (X_train['REGULARITY'])\n",
    "X_test['calls_density'] = X_test['total_calls'] / (X_test['REGULARITY'])\n",
    "X_val['calls_density'] = X_val['total_calls'] / (X_val['REGULARITY'])\n",
    "\n",
    "# 6.\n",
    "#X_train['arpu_last_to_average'] = X_train['ARPU_SEGMENT'] / (X_train['REVENUE'] + 0.0001)\n",
    "#X_test['arpu_last_to_average'] = X_test['ARPU_SEGMENT'] / (X_test['REVENUE'] + 0.0001)\n",
    "#X_val['arpu_last_to_average'] = X_val['ARPU_SEGMENT'] / (X_val['REVENUE'] + 0.0001)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Scaling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "\n",
    "#Standardize numeric columns # 13 шт.\n",
    "num_cols = ['MONTANT', 'FREQUENCE_RECH', 'REVENUE', 'ARPU_SEGMENT', 'FREQUENCE',\n",
    "       'DATA_VOLUME', 'ON_NET', 'ORANGE', 'TIGO',\n",
    "       'REGULARITY', 'FREQ_TOP_PACK', 'total_calls', 'MONTANT/FREQUENCE_RECH',\n",
    "           'remains', 'activ_density']\n",
    "\n",
    "scaler = StandardScaler()\n",
    "min_max_scaler = MinMaxScaler()\n",
    "\n",
    "X_train[num_cols] = scaler.fit_transform(X_train[num_cols])\n",
    "X_test[num_cols] = scaler.transform(X_test[num_cols])\n",
    "X_val[num_cols] = scaler.transform(X_val[num_cols])\n",
    "\n",
    "#X_train[num_cols] = min_max_scaler.fit_transform(X_train[num_cols])\n",
    "#X_test[num_cols] = min_max_scaler.transform(X_test[num_cols])\n",
    "#X_val[num_cols] = min_max_scaler.transform(X_val[num_cols])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### View from PCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "\n",
    "pca = PCA(svd_solver='full').fit(X_train.values)\n",
    "fig, ax1 = plt.subplots(1, 1, figsize=(8, 5))\n",
    "\n",
    "plt_grid = np.linspace(1, 101, 101)\n",
    "ax1.plot(np.arange(X_train.shape[1]) + 1, np.cumsum(pca.explained_variance_ratio_.tolist()))\n",
    "\n",
    "ax1.set_xlabel('Number of components', fontsize=14)\n",
    "ax1.set_ylabel('Cumulative explained variance', fontsize=14)\n",
    "ax1.grid(ls=':')\n",
    "ax1.set_xlim(0, 18)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'Principal component 2')"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "X_train_pca = pca.fit_transform(X_train)\n",
    "\n",
    "plt.figure(figsize=(8, 6))\n",
    "plt.scatter(X_train_pca[:10000, 0], X_train_pca[:10000, 1], c=y_train[:10000])\n",
    "plt.xlabel('Principal component 1')\n",
    "plt.ylabel('Principal component 2')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   33.5s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  2.5min\n",
      "[Parallel(n_jobs=-1)]: Done 200 out of 200 | elapsed:  2.7min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    2.5s\n",
      "[Parallel(n_jobs=8)]: Done 200 out of 200 | elapsed:    2.6s finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.6s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    2.6s\n",
      "[Parallel(n_jobs=8)]: Done 200 out of 200 | elapsed:    2.8s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.8503717802898416\n",
      "Recall: 0.31303013300769\n",
      "F1 Score: 0.4396875434611283\n",
      "AUC: 0.8986294701200582\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve\n",
    "from sklearn.metrics import auc\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix, recall_score, precision_recall_curve, f1_score\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "##RandomForestClassifier Model\n",
    "random_forest = RandomForestClassifier(bootstrap=True,criterion = \"gini\",\n",
    "                              n_jobs=-1,\n",
    "                              max_depth=7,\n",
    "                              n_estimators=200,\n",
    "                              random_state=1,\n",
    "                              verbose=True)\n",
    "\n",
    "#Fit model on data\n",
    "random_forest_model = random_forest.fit(X_train,y_train)\n",
    "\n",
    "#Predict on the X_test data \n",
    "random_forest_pred = random_forest_model.predict(X_test)\n",
    "random_forest_pred_proba = random_forest_model.predict_proba(X_test)[:,-1]\n",
    "\n",
    "print(\"Accuracy: {}\".format(accuracy_score(y_test, random_forest_pred)))\n",
    "print(\"Recall: {}\".format(recall_score(y_test, random_forest_pred)))\n",
    "print(\"F1 Score: {}\".format(f1_score(y_test, random_forest_pred)))\n",
    "\n",
    "fpr, tpr, thresholds = roc_curve(y_test.values, random_forest_pred_proba)\n",
    "print('AUC: {}'.format(auc(fpr, tpr)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Random Forest with Hyperparameter Selection using Validation Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 1 folds for each of 96 candidates, totalling 96 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed: 25.7min\n",
      "[Parallel(n_jobs=-1)]: Done  96 out of  96 | elapsed: 73.8min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'max_depth': 7, 'max_samples': 0.6, 'n_estimators': 300} 0.7414081595147528\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve\n",
    "from sklearn.metrics import auc\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix, recall_score, precision_recall_curve, f1_score\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "from sklearn.model_selection import PredefinedSplit\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import make_scorer\n",
    "\n",
    "\n",
    "\n",
    "# компановка валидационной и обучающей выборки для использования GridSearch\n",
    "X_train_and_val = pd.concat([X_train, X_val], axis=0)\n",
    "y_train_and_val = pd.concat([y_train, y_val], axis=0)\n",
    "\n",
    "# 0 для валидации, -1 для обучения\n",
    "train_and_val_flags = np.ones(X_train_and_val.shape[0])\n",
    "train_and_val_flags = -train_and_val_flags\n",
    "train_and_val_flags[:X_train.shape[0]] = 0\n",
    "\n",
    "ps = PredefinedSplit(test_fold=train_and_val_flags)\n",
    "\n",
    "\n",
    "# скоринг по AUC (целевой метрике)\n",
    "def AUC(y, y_pred):\n",
    "    fpr, tpr, thresholds = roc_curve(y.values, y_pred)\n",
    "    return auc(fpr, tpr)\n",
    "auc_scorer = make_scorer(AUC)\n",
    "\n",
    "\n",
    "\n",
    "RF_param_grid = {'n_estimators': np.linspace(150, 400, 6, dtype=int), \n",
    "                 'max_samples': np.linspace(0.4, 1, 4),\n",
    "                 'max_depth': np.linspace(7, 15, 4, dtype=int)}\n",
    "\n",
    "RF_selection_model = GridSearchCV(RandomForestClassifier(n_jobs=-1, random_state=1, verbose=True),\n",
    "                             RF_param_grid, scoring=auc_scorer,\n",
    "                             n_jobs=-1, refit=False, cv=ps, verbose=True)\n",
    "\n",
    "RF_selection_model.fit(X_train_and_val, y_train_and_val)\n",
    "print(RF_selection_model.best_params_, RF_selection_model.best_score_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   33.0s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  2.5min\n",
      "[Parallel(n_jobs=-1)]: Done 300 out of 300 | elapsed:  4.0min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    2.4s\n",
      "[Parallel(n_jobs=8)]: Done 300 out of 300 | elapsed:    3.9s finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    2.5s\n",
      "[Parallel(n_jobs=8)]: Done 300 out of 300 | elapsed:    4.0s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.8503857075431551\n",
      "Recall: 0.3124278028977854\n",
      "F1 Score: 0.4392360204856942\n",
      "AUC: 0.8986907102361222\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve\n",
    "from sklearn.metrics import auc\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix, recall_score, precision_recall_curve, f1_score\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "from sklearn.model_selection import PredefinedSplit\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import make_scorer\n",
    "\n",
    "\n",
    "X_train_and_val = pd.concat([X_train, X_val], axis=0)\n",
    "y_train_and_val = pd.concat([y_train, y_val], axis=0)\n",
    "#Predict on the X_test data\n",
    "#RF_selected_model = RandomForestClassifier(n_estimators=RF_selection_model.best_params_['n_estimators'], \n",
    "#                                          max_samples=RF_selection_model.best_params_['max_samples'],\n",
    "#                                          max_depth=RF_selection_model.best_params_['max_depth'],\n",
    "#                                          n_jobs=-1, random_state=1, verbose=True)\n",
    "\n",
    "RF_selected_model = RandomForestClassifier(n_estimators=300, \n",
    "                                          max_samples=0.6,\n",
    "                                          max_depth=7,\n",
    "                                          n_jobs=-1, random_state=1, verbose=True)\n",
    "\n",
    "\n",
    "RF_selected_model.fit(X_train_and_val, y_train_and_val)\n",
    "RF_selected_pred = RF_selected_model.predict(X_test)\n",
    "RF_selected_pred_proba = RF_selected_model.predict_proba(X_test)[:,-1]\n",
    "\n",
    "print(\"Accuracy: {}\".format(accuracy_score(y_test, RF_selected_pred)))\n",
    "print(\"Recall: {}\".format(recall_score(y_test, RF_selected_pred)))\n",
    "print(\"F1 Score: {}\".format(f1_score(y_test, RF_selected_pred)))\n",
    "\n",
    "fpr, tpr, thresholds = roc_curve(y_test.values, RF_selected_pred_proba)\n",
    "print('AUC: {}'.format(auc(fpr, tpr)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Take 10 best Random Forest models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['mean_fit_time', 'std_fit_time', 'mean_score_time', 'std_score_time',\n",
       "       'param_max_depth', 'param_max_samples', 'param_n_estimators', 'params',\n",
       "       'split0_test_score', 'mean_test_score', 'std_test_score',\n",
       "       'rank_test_score'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "RF_models = pd.DataFrame(RF_selection_model.cv_results_)\n",
    "RF_models.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/10 [00:00<?, ?it/s][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   23.0s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  2.1min\n",
      "[Parallel(n_jobs=-1)]: Done 300 out of 300 | elapsed:  3.2min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    2.5s\n",
      "[Parallel(n_jobs=8)]: Done 300 out of 300 | elapsed:    4.0s finished\n",
      " 10%|█         | 1/10 [03:17<29:33, 197.02s/it][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   22.6s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  1.9min\n",
      "[Parallel(n_jobs=-1)]: Done 200 out of 200 | elapsed:  2.0min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    2.4s\n",
      "[Parallel(n_jobs=8)]: Done 200 out of 200 | elapsed:    2.6s finished\n",
      " 20%|██        | 2/10 [05:21<23:22, 175.31s/it][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   24.2s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  2.1min\n",
      "[Parallel(n_jobs=-1)]: Done 300 out of 300 | elapsed:  3.4min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.7s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    3.2s\n",
      "[Parallel(n_jobs=8)]: Done 300 out of 300 | elapsed:    5.4s finished\n",
      " 30%|███       | 3/10 [08:49<21:34, 184.96s/it][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   23.8s\n",
      "[Parallel(n_jobs=-1)]: Done 150 out of 150 | elapsed:  1.6min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.7s\n",
      "[Parallel(n_jobs=8)]: Done 150 out of 150 | elapsed:    2.6s finished\n",
      " 40%|████      | 4/10 [10:27<15:54, 159.11s/it][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   19.6s\n",
      "[Parallel(n_jobs=-1)]: Done 150 out of 150 | elapsed:  1.3min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done 150 out of 150 | elapsed:    2.1s finished\n",
      " 50%|█████     | 5/10 [11:48<11:17, 135.50s/it][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   25.9s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  2.0min\n",
      "[Parallel(n_jobs=-1)]: Done 250 out of 250 | elapsed:  2.7min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.7s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    3.2s\n",
      "[Parallel(n_jobs=8)]: Done 250 out of 250 | elapsed:    4.3s finished\n",
      " 60%|██████    | 6/10 [14:35<09:39, 144.88s/it][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   25.5s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  2.0min\n",
      "[Parallel(n_jobs=-1)]: Done 200 out of 200 | elapsed:  2.2min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.7s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    3.2s\n",
      "[Parallel(n_jobs=8)]: Done 200 out of 200 | elapsed:    3.4s finished\n",
      " 70%|███████   | 7/10 [16:49<07:05, 141.76s/it][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   25.4s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  2.0min\n",
      "[Parallel(n_jobs=-1)]: Done 350 out of 350 | elapsed:  3.7min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.7s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    3.2s\n",
      "[Parallel(n_jobs=8)]: Done 350 out of 350 | elapsed:    6.0s finished\n",
      " 80%|████████  | 8/10 [20:39<05:36, 168.10s/it][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   24.0s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  1.9min\n",
      "[Parallel(n_jobs=-1)]: Done 400 out of 400 | elapsed:  4.1min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.7s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    3.2s\n",
      "[Parallel(n_jobs=8)]: Done 400 out of 400 | elapsed:    6.8s finished\n",
      " 90%|█████████ | 9/10 [24:53<03:13, 193.84s/it][Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   17.2s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  1.4min\n",
      "[Parallel(n_jobs=-1)]: Done 400 out of 400 | elapsed:  2.9min finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    2.4s\n",
      "[Parallel(n_jobs=8)]: Done 400 out of 400 | elapsed:    5.2s finished\n",
      "100%|██████████| 10/10 [27:54<00:00, 167.45s/it]\n"
     ]
    }
   ],
   "source": [
    "from tqdm import tqdm\n",
    "\n",
    "best_RF_probs = []\n",
    "for i in tqdm(range(1, 11)):\n",
    "    params = RF_models[RF_models['rank_test_score'] == i]['params'].values[0]\n",
    "    RF_model = RandomForestClassifier(n_estimators=params['n_estimators'], \n",
    "                                      max_samples=params['max_samples'],\n",
    "                                      max_depth=params['max_depth'],\n",
    "                                      n_jobs=-1, random_state=1, verbose=True)\n",
    "    RF_model.fit(X_train, y_train)\n",
    "    best_RF_probs.append(RF_model.predict_proba(X_test)[:,-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.898801781483697\n"
     ]
    }
   ],
   "source": [
    "average_prediction_proba_RF = np.mean(np.array(best_RF_probs), axis=0)\n",
    "fpr, tpr, thresholds = roc_curve(y_test.values, average_prediction_proba_RF)\n",
    "print('AUC: {}'.format(auc(fpr, tpr)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. SVM models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import roc_curve\n",
    "from sklearn.metrics import auc\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix, recall_score, precision_recall_curve, f1_score\n",
    "\n",
    "from sklearn.svm import SVC\n",
    "\n",
    "from sklearn.model_selection import PredefinedSplit\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import make_scorer\n",
    "\n",
    "\n",
    "# компановка валидационной и обучающей выборки для использования GridSearch\n",
    "X_train_and_val = pd.concat([X_train, X_val], axis=0)\n",
    "y_train_and_val = pd.concat([y_train, y_val], axis=0)\n",
    "\n",
    "# 0 для валидации, -1 для обучения\n",
    "train_and_val_flags = np.ones(X_train_and_val.shape[0])\n",
    "train_and_val_flags = -train_and_val_flags\n",
    "train_and_val_flags[:X_train.shape[0]] = 0\n",
    "\n",
    "ps = PredefinedSplit(test_fold=train_and_val_flags)\n",
    "\n",
    "\n",
    "# скоринг по AUC (целевой метрике)\n",
    "def AUC(y, y_pred):\n",
    "    fpr, tpr, thresholds = roc_curve(y.values, y_pred)\n",
    "    return auc(fpr, tpr)\n",
    "auc_scorer = make_scorer(AUC)\n",
    "\n",
    "\n",
    "# degree (poly)\n",
    "SVM_param_grid = {'C': np.logspace(-3, 3, 6),\n",
    "                  'kernel': ['poly', 'rbf', 'sigmoid'],\n",
    "                  'degree': np.linspace(3, 6, 4, dtype=int)}\n",
    "\n",
    "SVM_selection_model = GridSearchCV(SVC(random_state=1, verbose=True),\n",
    "                             SVM_param_grid, scoring=auc_scorer,\n",
    "                             n_jobs=-1, refit=False, cv=ps, verbose=True)\n",
    "\n",
    "SVM_selection_model.fit(X_train_and_val, y_train_and_val)\n",
    "print(SVM_selection_model.best_params_, SVM_selection_model.best_score_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Predict on the X_test data\n",
    "SVM_selected_model = SVC(C=SVM_selection_model.best_params_['C'],\n",
    "                         kernel=SVM_selection_model.best_params_['kernel'],\n",
    "                         degree=SVM_selection_model.best_params_['degree'],\n",
    "                         random_state=1, verbose=True)\n",
    "\n",
    "SVM_selected_model.fit(X_train_and_val, y_train_and_val)\n",
    "SVM_selected_pred = SVM_selected_model.predict(X_test)\n",
    "SVM_selected_pred_proba = SVM_selected_model.predict_proba(X_test)[:,-1]\n",
    "\n",
    "print(\"Accuracy: {}\".format(accuracy_score(y_test, SVM_selected_pred)))\n",
    "print(\"Recall: {}\".format(recall_score(y_test, SVM_selected_pred)))\n",
    "print(\"F1 Score: {}\".format(f1_score(y_test, SVM_selected_pred)))\n",
    "\n",
    "fpr, tpr, thresholds = roc_curve(y_test.values, SVM_selected_pred_proba)\n",
    "print('AUC: {}'.format(auc(fpr, tpr)))"
   ]
  },
  {
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}