notebook/Fraud_Detection.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fraud Prediction Using Skewed Data\n",
    "\n",
    "### How do we achieve good predictive power in the statistical model? Can sampling techniques help? Let's find out!\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Setup\n",
    "To prepare your environment, you need to install some packages.\n",
    "\n",
    "### 1.1 Install the necessary packages\n",
    "\n",
    "You need the latest versions of these packages:<br>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np \n",
    "import pandas as pd \n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Read the Data & convert it into Dataframe\n",
    "Click on Insert to code and then select Insert pandas DataFrame in the below empty cell."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>Time</th>\n",
       "      <th>V1</th>\n",
       "      <th>V2</th>\n",
       "      <th>V3</th>\n",
       "      <th>V4</th>\n",
       "      <th>V5</th>\n",
       "      <th>V6</th>\n",
       "      <th>V7</th>\n",
       "      <th>V8</th>\n",
       "      <th>V9</th>\n",
       "      <th>...</th>\n",
       "      <th>V21</th>\n",
       "      <th>V22</th>\n",
       "      <th>V23</th>\n",
       "      <th>V24</th>\n",
       "      <th>V25</th>\n",
       "      <th>V26</th>\n",
       "      <th>V27</th>\n",
       "      <th>V28</th>\n",
       "      <th>Amount</th>\n",
       "      <th>Class</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>10</td>\n",
       "      <td>0.385</td>\n",
       "      <td>0.616</td>\n",
       "      <td>-0.874</td>\n",
       "      <td>-0.094</td>\n",
       "      <td>2.925</td>\n",
       "      <td>3.317</td>\n",
       "      <td>0.470</td>\n",
       "      <td>0.538</td>\n",
       "      <td>-0.559</td>\n",
       "      <td>...</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.238</td>\n",
       "      <td>0.009</td>\n",
       "      <td>0.997</td>\n",
       "      <td>-0.767</td>\n",
       "      <td>-0.492</td>\n",
       "      <td>0.042</td>\n",
       "      <td>-0.054</td>\n",
       "      <td>9.99</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>12</td>\n",
       "      <td>-0.752</td>\n",
       "      <td>0.345</td>\n",
       "      <td>2.057</td>\n",
       "      <td>-1.469</td>\n",
       "      <td>-1.158</td>\n",
       "      <td>-0.078</td>\n",
       "      <td>-0.609</td>\n",
       "      <td>0.004</td>\n",
       "      <td>-0.436</td>\n",
       "      <td>...</td>\n",
       "      <td>0.500</td>\n",
       "      <td>1.354</td>\n",
       "      <td>-0.257</td>\n",
       "      <td>-0.065</td>\n",
       "      <td>-0.039</td>\n",
       "      <td>-0.087</td>\n",
       "      <td>-0.181</td>\n",
       "      <td>0.129</td>\n",
       "      <td>15.99</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>25</td>\n",
       "      <td>1.114</td>\n",
       "      <td>0.086</td>\n",
       "      <td>0.494</td>\n",
       "      <td>1.336</td>\n",
       "      <td>-0.300</td>\n",
       "      <td>-0.011</td>\n",
       "      <td>-0.119</td>\n",
       "      <td>0.189</td>\n",
       "      <td>0.206</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.053</td>\n",
       "      <td>-0.005</td>\n",
       "      <td>-0.031</td>\n",
       "      <td>0.198</td>\n",
       "      <td>0.565</td>\n",
       "      <td>-0.338</td>\n",
       "      <td>0.029</td>\n",
       "      <td>0.004</td>\n",
       "      <td>4.45</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>33</td>\n",
       "      <td>-0.936</td>\n",
       "      <td>0.170</td>\n",
       "      <td>2.746</td>\n",
       "      <td>-1.078</td>\n",
       "      <td>-0.306</td>\n",
       "      <td>0.012</td>\n",
       "      <td>-0.296</td>\n",
       "      <td>0.403</td>\n",
       "      <td>-0.040</td>\n",
       "      <td>...</td>\n",
       "      <td>0.401</td>\n",
       "      <td>1.065</td>\n",
       "      <td>-0.158</td>\n",
       "      <td>0.296</td>\n",
       "      <td>-0.259</td>\n",
       "      <td>0.754</td>\n",
       "      <td>0.047</td>\n",
       "      <td>0.094</td>\n",
       "      <td>9.10</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>35</td>\n",
       "      <td>1.199</td>\n",
       "      <td>0.130</td>\n",
       "      <td>0.864</td>\n",
       "      <td>1.003</td>\n",
       "      <td>-0.784</td>\n",
       "      <td>-0.885</td>\n",
       "      <td>-0.041</td>\n",
       "      <td>-0.208</td>\n",
       "      <td>0.392</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.042</td>\n",
       "      <td>0.198</td>\n",
       "      <td>-0.033</td>\n",
       "      <td>1.013</td>\n",
       "      <td>0.559</td>\n",
       "      <td>0.402</td>\n",
       "      <td>-0.006</td>\n",
       "      <td>0.018</td>\n",
       "      <td>0.99</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 31 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Time     V1     V2     V3     V4     V5     V6     V7     V8     V9  ...    \\\n",
       "0    10  0.385  0.616 -0.874 -0.094  2.925  3.317  0.470  0.538 -0.559  ...     \n",
       "1    12 -0.752  0.345  2.057 -1.469 -1.158 -0.078 -0.609  0.004 -0.436  ...     \n",
       "2    25  1.114  0.086  0.494  1.336 -0.300 -0.011 -0.119  0.189  0.206  ...     \n",
       "3    33 -0.936  0.170  2.746 -1.078 -0.306  0.012 -0.296  0.403 -0.040  ...     \n",
       "4    35  1.199  0.130  0.864  1.003 -0.784 -0.885 -0.041 -0.208  0.392  ...     \n",
       "\n",
       "     V21    V22    V23    V24    V25    V26    V27    V28  Amount  Class  \n",
       "0  0.050  0.238  0.009  0.997 -0.767 -0.492  0.042 -0.054    9.99      0  \n",
       "1  0.500  1.354 -0.257 -0.065 -0.039 -0.087 -0.181  0.129   15.99      0  \n",
       "2 -0.053 -0.005 -0.031  0.198  0.565 -0.338  0.029  0.004    4.45      0  \n",
       "3  0.401  1.065 -0.158  0.296 -0.259  0.754  0.047  0.094    9.10      0  \n",
       "4 -0.042  0.198 -0.033  1.013  0.559  0.402 -0.006  0.018    0.99      0  \n",
       "\n",
       "[5 rows x 31 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Assign a new name for the dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "'''Rename the dataframe to df'''\n",
    "\n",
    "df = df_data_2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Descriptive statistics on the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time\n",
      "count     56874.000000\n",
      "mean      94995.469934\n",
      "std       47535.817923\n",
      "min          10.000000\n",
      "25%       54380.250000\n",
      "50%       84664.500000\n",
      "75%      139664.500000\n",
      "max      172782.000000\n",
      "Name: Time, dtype: float64\n",
      "V1\n",
      "count     56874.000000\n",
      "mean      94995.469934\n",
      "std       47535.817923\n",
      "min          10.000000\n",
      "25%       54380.250000\n",
      "50%       84664.500000\n",
      "75%      139664.500000\n",
      "max      172782.000000\n",
      "Name: Time, dtype: float64\n",
      "Amount\n",
      "count    56874.000000\n",
      "mean        88.264681\n",
      "std        233.922284\n",
      "min          0.000000\n",
      "25%          5.740000\n",
      "50%         22.205000\n",
      "75%         78.600000\n",
      "max       7862.390000\n",
      "Name: Amount, dtype: float64\n",
      "Class\n",
      "count    56874.000000\n",
      "mean         0.001793\n",
      "std          0.042311\n",
      "min          0.000000\n",
      "25%          0.000000\n",
      "50%          0.000000\n",
      "75%          0.000000\n",
      "max          1.000000\n",
      "Name: Class, dtype: float64\n",
      "2\n",
      "ALL\n",
      "                Time            V1            V2            V3            V4  \\\n",
      "count   56874.000000  56874.000000  56874.000000  56874.000000  56874.000000   \n",
      "mean    94995.469934      0.010646      0.004147     -0.003954      0.002830   \n",
      "std     47535.817923      1.933044      1.604704      1.514694      1.412792   \n",
      "min        10.000000    -41.929000    -47.430000    -30.177000     -5.683000   \n",
      "25%     54380.250000     -0.915000     -0.606000     -0.893000     -0.846000   \n",
      "50%     84664.500000      0.027000      0.060000      0.181000     -0.023000   \n",
      "75%    139664.500000      1.320000      0.805000      1.021000      0.748000   \n",
      "max    172782.000000      2.455000     22.058000      9.383000     11.865000   \n",
      "\n",
      "                 V5            V6            V7            V8            V9  \\\n",
      "count  56874.000000  56874.000000  56874.000000  56874.000000  56874.000000   \n",
      "mean      -0.001663      0.007861     -0.004580      0.003089      0.000165   \n",
      "std        1.356933      1.329296      1.211257      1.193217      1.108262   \n",
      "min      -22.065000    -23.497000    -43.557000    -50.943000    -13.320000   \n",
      "25%       -0.705000     -0.767000     -0.560000     -0.209000     -0.648000   \n",
      "50%       -0.061000     -0.269000      0.036500      0.023000     -0.052000   \n",
      "75%        0.607000      0.407000      0.564000      0.332000      0.596000   \n",
      "max       34.099000     16.356000     24.665000     18.749000     15.595000   \n",
      "\n",
      "           ...                V21           V22           V23           V24  \\\n",
      "count      ...       56874.000000  56874.000000  56874.000000  56874.000000   \n",
      "mean       ...           0.000675     -0.000240     -0.002827     -0.000386   \n",
      "std        ...           0.725244      0.720987      0.603752      0.603345   \n",
      "min        ...         -22.798000     -8.887000    -21.304000     -2.823000   \n",
      "25%        ...          -0.229000     -0.542000     -0.161000     -0.352000   \n",
      "50%        ...          -0.029000      0.006000     -0.011000      0.040000   \n",
      "75%        ...           0.184000      0.525000      0.149000      0.439000   \n",
      "max        ...          27.203000      8.316000     18.364000      4.016000   \n",
      "\n",
      "                V25           V26           V27           V28        Amount  \\\n",
      "count  56874.000000  56874.000000  56874.000000  56874.000000  56874.000000   \n",
      "mean      -0.000068     -0.000414      0.002503      0.001734     88.264681   \n",
      "std        0.522217      0.481846      0.399660      0.311913    233.922284   \n",
      "min       -5.785000     -2.069000    -22.566000    -11.711000      0.000000   \n",
      "25%       -0.316000     -0.327000     -0.070000     -0.053000      5.740000   \n",
      "50%        0.017000     -0.055000      0.002000      0.011000     22.205000   \n",
      "75%        0.350000      0.239000      0.092000      0.079000     78.600000   \n",
      "max        5.474000      3.463000      8.254000     16.130000   7862.390000   \n",
      "\n",
      "              Class  \n",
      "count  56874.000000  \n",
      "mean       0.001793  \n",
      "std        0.042311  \n",
      "min        0.000000  \n",
      "25%        0.000000  \n",
      "50%        0.000000  \n",
      "75%        0.000000  \n",
      "max        1.000000  \n",
      "\n",
      "[8 rows x 31 columns]\n"
     ]
    }
   ],
   "source": [
    "print(\"Time\")\n",
    "print(df.Time.describe())\n",
    "print(\"V1\")\n",
    "print(df.Time.describe())\n",
    "print(\"Amount\")\n",
    "print(df.Amount.describe())\n",
    "print(\"Class\")\n",
    "print(df.Class.describe())\n",
    "print(df.Class.nunique())\n",
    "print(\"ALL\")\n",
    "print(df.describe())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. Use Histogram to visualize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2accb2b48550>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2acc82b57d68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.Time.hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2accb4b51978>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2acc82b57dd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.Class.hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([<matplotlib.axes._subplots.AxesSubplot object at 0x2accb4be1320>,\n",
       "       <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4c1c048>], dtype=object)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZEAAAEWCAYAAACnlKo3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4wLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvpW3flQAAHDFJREFUeJzt3X2QXXWd5/H3x0QUUR4NWUzA4Bod0VkQeyE71G45RCCoZZgpGHEeiFbcbDm4gztTq9G1lvJpC7Z2B6VK3U1JJDhqRJQlq2hIRZgtdxFong0BEiMPbSCJkxBBZ9DgZ/84v5abzu2nk9v3nG4+r6que+/vnHP7e0/n5HMefvd3ZJuIiIg6XtR0ARERMX0lRCIioraESERE1JYQiYiI2hIiERFRW0IkIiJqS4hERERtCZEZTNLRkq6X9EtJj0r606ZrimiCpA9KGpT0rKSrm65nJpnddAExpT4P/BqYC5wCfFfSvbY3NVtWRN9tBz4NnAMc2nAtM4ryjfWZSdJhwB7gTbYfLm1fAX5me2WjxUU0RNKngfm239t0LTNFTmfNXK8DnhsOkOJe4I0N1RMRM1BCZOZ6ObB3RNte4BUN1BIRM1RCZOZ6Bjh8RNvhwNMN1BIRM1RCZOZ6GJgtaWFH28lALqpHRM8kRGYo278Evg18UtJhks4AlgJfabayiP6TNFvSS4FZwCxJL5WU3qk9kBCZ2f6SqjvjTuDrwAfSvTdeoD4O/COwEvjz8vzjjVY0Q6SLb0RE1JYjkYiIqC0hEhERtSVEIiKitoRIRETUlhCJiIjapm0/6Ve+8pVesGBB02XEDHPnnXf+3PacpuuYjGwLMRUmui1M2xBZsGABg4ODTZcRM4ykR5uuYbKyLcRUmOi2kNNZERFRW0IkIiJqS4hERERtCZGIiKgtIRIREbUlRCIioraESERE1JYQiYiI2qbtlw1Hs2Dld2st98hl7+hxJRHNq7M9ZFuIyciRSERE1JYQiYiI2hIiERFRW0IkIiJqS4hERERtCZGIhkj6D5I2SfqxpK9LeqmkEyXdJmmLpG9IOqTpOiPGkhCJaICkecBfAQO23wTMAi4ELgeusL0Q2AMsb67KiPElRCKaMxs4VNJs4GXAE8CZwHVl+hrgvIZqi5iQhEhEA2z/DPhvwGNU4bEXuBN4yva+MtsQMK+ZCiMmJiES0QBJRwFLgROBVwGHAed2mdWjLL9C0qCkwV27dk1doRHjSIhENONtwE9t77L9G+DbwB8AR5bTWwDzge3dFra9yvaA7YE5c+b0p+KILhIiEc14DFgk6WWSBCwGHgBuBs4v8ywDbmiovogJSYhENMD2bVQX0O8C7qfaFlcBHwH+WtJW4BjgqsaKjJiAGTeKb8R0YftS4NIRzduA0xooJ6KWHIlERERtCZGIiKgtIRIREbVNKEQkPSLpfkn3SBosbUdL2lDG+NlQ+r2jypWStkq6T9KpHe+zrMy/RdKyjva3lPffWpZVrz9oRET03mSORP7Q9im2B8rrlcDGMsbPxvIaqi9MLSw/K4AvQhU6VBcRT6e6cHjpcPCUeVZ0LLek9ieKiIi+OZjTWUupxvaB/cf4WQpc48qPqL48dRxwDrDB9m7be4ANwJIy7XDbt9o2cA0ZLygiYlqYaIgYuEnSnZJWlLa5tp8AKI/HlvZ5wOMdyw6P/zNW+1CX9oiIaLmJfk/kDNvbJR0LbJD04Bjzdrue4RrtB75xFWArAE444YSxK46IiCk3oSMR29vL407geqprGjvKqSjK484y+xBwfMfiw+P/jNU+v0t7tzoyXlBERIuMGyKSDpP0iuHnwNnAj4F1VGP7wP5j/KwDLiq9tBYBe8vprvXA2ZKOKhfUzwbWl2lPS1pUemVdRMYLioiYFiZyOmsucH3pdTsb+Jrt70u6A7hW0nKqweQuKPPfCLwd2Ar8CngfgO3dkj4F3FHm+6Tt3eX5B4CrgUOB75WfiIhouXFDxPY24OQu7f9ANfLoyHYDF4/yXquB1V3aB4E3TaDeiIhokXxjPSIiakuIREREbQmRiIioLSESERG1JUQiGiDp9WVA0+GfX0j60GgDm0a0VUIkogG2HyoDmp4CvIWqO/z1jD6waUQrJUQimrcY+IntRxl9YNOIVkqIRDTvQuDr5floA5vuR9IKSYOSBnft2tWnMiMOlBCJaJCkQ4B3Ad+czHIZRy7aIiES0axzgbts7yivRxvYNKKVEiIRzXoPz5/KgtEHNo1opYRIREMkvQw4C/h2R/NlwFmStpRplzVRW8RETfSmVBHRY7Z/BRwzoq3rwKYRbZUjkYiIqC0hEhERtSVEIiKitoRIRETUlhCJiIjaEiIREVFbQiQiImpLiERERG0JkYiIqC0hEhERtSVEIiKitoRIRETUlhCJiIjaEiIREVFbQiQiImqbcIhImiXpbknfKa9PlHSbpC2SvlHuFY2kl5TXW8v0BR3v8dHS/pCkczral5S2rZJW9u7jRbSXpCMlXSfpQUmbJf0rSUdL2lC2qw2Sjmq6zoixTOZI5BJgc8fry4ErbC8E9gDLS/tyYI/t1wJXlPmQdBJwIfBGYAnwhRJMs4DPU91r+iTgPWXeiJnuc8D3bf8ecDLV9rUS2Fi2q43ldURrTShEJM0H3gF8qbwWcCZwXZllDXBeeb60vKZMX1zmXwqstf2s7Z8CW4HTys9W29ts/xpYW+aNmLEkHQ78G+AqANu/tv0U+28/ndtVRCtN9Ejks8CHgd+W18cAT9neV14PAfPK83nA4wBl+t4y/+/aRywzWvsBJK2QNChpcNeuXRMsPaKVXgPsAr5cThN/SdJhwFzbTwCUx2ObLDJiPOOGiKR3Ajtt39nZ3GVWjzNtsu0HNtqrbA/YHpgzZ84YVUe03mzgVOCLtt8M/JJJnLrKDlW0xUSORM4A3iXpEapTTWdSHZkcKWl2mWc+sL08HwKOByjTjwB2d7aPWGa09oiZbAgYsn1beX0dVajskHQcQHnc2W3h7FBFW4wbIrY/anu+7QVUF8Z/YPvPgJuB88tsy4AbyvN15TVl+g9su7RfWHpvnQgsBG4H7gAWlt5eh5Tfsa4nny6ipWw/CTwu6fWlaTHwAPtvP53bVUQrzR5/llF9BFgr6dPA3ZQLhOXxK5K2Uh2BXAhge5Oka6k2lH3AxbafA5D0QWA9MAtYbXvTQdQVMV38e+CrZedpG/A+qh27ayUtBx4DLmiwvohxTSpEbN8C3FKeb6PqWTVynn9ilH/4tj8DfKZL+43AjZOpJWK6s30PMNBl0uJ+1xJRV76xHhERtSVEIiKitoRIRETUlhCJiIjaEiIREVFbQiQiImpLiERERG0JkYiIqC0hEhERtSVEIiKitoRIRETUlhCJiIjaEiIREVFbQiQiImpLiERERG0JkYiIqO1g7mwYEQdB0iPA08BzwD7bA5KOBr4BLAAeAf7E9p6maowYT45EIpr1h7ZPsT18h8OVwEbbC4GN5XVEayVEItplKbCmPF8DnNdgLRHjSohENMfATZLulLSitM21/QRAeTy224KSVkgalDS4a9euPpUbcaBcE4lozhm2t0s6Ftgg6cGJLmh7FbAKYGBgwFNVYMR4ciQS0RDb28vjTuB64DRgh6TjAMrjzuYqjBhfQiSiAZIOk/SK4efA2cCPgXXAsjLbMuCGZiqMmJiczopoxlzgeklQbYdfs/19SXcA10paDjwGXNBgjRHjSohENMD2NuDkLu3/ACzuf0UR9eR0VkRE1JYQiYiI2hIiERFR27ghIumlkm6XdK+kTZI+UdpPlHSbpC2SviHpkNL+kvJ6a5m+oOO9PlraH5J0Tkf7ktK2VVKGeYiImCYmciTyLHCm7ZOBU4AlkhYBlwNXlDF+9gDLy/zLgT22XwtcUeZD0knAhcAbgSXAFyTNkjQL+DxwLnAS8J4yb0REtNy4IeLKM+Xli8uPgTOB60p75xg/nWP/XAcsVtWPcSmw1vaztn8KbKX6ctVpwFbb22z/Glhb5o2IiJab0DWRcsRwD9W3ZzcAPwGesr2vzDIEzCvP5wGPA5Tpe4FjOttHLDNae7c6Ml5QRESLTChEbD9n+xRgPtWRwxu6zVYeNcq0ybZ3q2OV7QHbA3PmzBm/8IiImFKT6p1l+yngFmARcKSk4S8rzge2l+dDwPEAZfoRwO7O9hHLjNYeEREtN5HeWXMkHVmeHwq8DdgM3AycX2brHOOnc+yf84Ef2HZpv7D03joRWAjcDtwBLCy9vQ6huvi+rhcfLiIiptZEhj05DlhTelG9CLjW9nckPQCslfRp4G7gqjL/VcBXJG2lOgK5EMD2JknXAg8A+4CLbT8HIOmDwHpgFrDa9qaefcKIiJgy44aI7fuAN3dp30Z1fWRk+z8xyqBxtj8DfKZL+43AjROoNyIiWiTfWI+IiNoSIhERUVtCJCIiakuIRDSofJH3bknfKa+7jkkX0VYJkYhmXULVZX7YaGPSRbRSQiSiIZLmA+8AvlRei9HHpItopYRIRHM+C3wY+G15fQyjj0kX0UoJkYgGSHonsNP2nZ3NXWbtOo5cBiONtkiIRDTjDOBdkh6huv3BmVRHJqONSbefDEYabZEQiWiA7Y/anm97AdXQQD+w/WeMPiZdRCslRCLa5SPAX5ex547h+THpIlppIgMwRsQUsn0L1S0WRh2TLqKtciQSERG1JUQiIqK2hEhERNSWEImIiNoSIhERUVtCJCIiakuIREREbQmRiIioLSESERG1JUQiIqK2hEhERNSWEImIiNoSIhERUVtCJCIiakuIREREbQmRiIiobdwQkXS8pJslbZa0SdIlpf1oSRskbSmPR5V2SbpS0lZJ90k6teO9lpX5t0ha1tH+Fkn3l2WulKSp+LARbSHppZJul3Rv2a4+UdpPlHRb2Ua+IemQpmuNGMtEjkT2AX9j+w3AIuBiSScBK4GNthcCG8trgHOBheVnBfBFqEIHuBQ4nerObZcOB0+ZZ0XHcksO/qNFtNqzwJm2TwZOAZZIWgRcDlxRtqs9wPIGa4wY17ghYvsJ23eV508Dm4F5wFJgTZltDXBeeb4UuMaVHwFHSjoOOAfYYHu37T3ABqoN5zjgcNu32jZwTcd7RcxIZft4prx8cfkxcCZwXWnv3K4iWmlS10QkLQDeDNwGzLX9BFRBAxxbZpsHPN6x2FBpG6t9qEt7t9+/QtKgpMFdu3ZNpvSI1pE0S9I9wE6qnaqfAE/Z3ldmybYQrTfhEJH0cuBbwIds/2KsWbu0uUb7gY32KtsDtgfmzJkzXskRrWb7OdunAPOpTvG+odtsoyybbSFaYUIhIunFVAHyVdvfLs07yqkoyuPO0j4EHN+x+Hxg+zjt87u0R7wg2H4KuIXqmuORkmaXSdkWovUm0jtLwFXAZtt/2zFpHTDcw2oZcENH+0Wll9YiYG853bUeOFvSUeWC+tnA+jLtaUmLyu+6qOO9ImYkSXMkHVmeHwq8jep6483A+WW2zu0qopVmjz8LZwB/Adxfzt8CfAy4DLhW0nLgMeCCMu1G4O3AVuBXwPsAbO+W9CngjjLfJ23vLs8/AFwNHAp8r/xEzGTHAWskzaLambvW9nckPQCslfRp4G6qHbiI1ho3RGz/kO7XLQAWd5nfwMWjvNdqYHWX9kHgTePVEjFT2L6PqpPKyPZtVNdHIqaFfGM9IiJqS4hERERtCZGIiKgtIRIREbUlRCIioraESERE1JYQiYiI2hIiERFRW0IkIiJqS4hERERtCZGIiKgtIRIREbUlRCIioraESERE1JYQiYiI2hIiEQ2QdLykmyVtlrRJ0iWl/WhJGyRtKY9HNV1rxFgSIhHN2Af8je03UN1b/WJJJwErgY22FwIby+uI1kqIRDTA9hO27yrPn6a6v/o8YCmwpsy2BjivmQojJiYhEtEwSQuobpV7GzDX9hNQBQ1wbHOVRYwvIRLRIEkvB74FfMj2Lyax3ApJg5IGd+3aNXUFRowjIRLREEkvpgqQr9r+dmneIem4Mv04YGe3ZW2vsj1ge2DOnDn9KTiii4RIRAMkCbgK2Gz7bzsmrQOWlefLgBv6XVvEZMxuuoCIF6gzgL8A7pd0T2n7GHAZcK2k5cBjwAUN1RcxIQmRiAbY/iGgUSYv7mctEQcjp7MiIqK2hEhERNSWEImIiNrGDRFJqyXtlPTjjrau4/uocqWkrZLuk3RqxzLLyvxbJC3raH+LpPvLMleWXisRETENTORI5GpgyYi20cb3ORdYWH5WAF+EKnSAS4HTgdOASzsGlvtimXd4uZG/KyIiWmrcELH9f4DdI5pHG99nKXCNKz8CjixfmDoH2GB7t+09wAZgSZl2uO1bbRu4howVFBExbdS9JjLa+D7zgMc75hsqbWO1D3Vpj4iIaaDXF9a7Xc9wjfbub57xgiIiWqVuiIw2vs8QcHzHfPOB7eO0z+/S3lXGC4qIaJe6ITLa+D7rgItKL61FwN5yums9cLako8oF9bOB9WXa05IWlV5ZF5GxgiIipo1xhz2R9HXgrcArJQ1R9bIabXyfG4G3A1uBXwHvA7C9W9KngDvKfJ+0PXyx/gNUPcAOBb5XfiIiYhoYN0Rsv2eUSQeM71N6WF08yvusBlZ3aR8E3jReHRER0T75xnpERNSWEImIiNoSIhERUVtCJCIiakuIRDRgMgObRrRZQiSiGVcz8YFNI1orIRLRgEkObBrRWgmRiPYYbWDTA2QcuWiLhEjENJRx5KItEiIR7THawKYRrZUQiWiP0QY2jWithEhEA8rAprcCr5c0VAYzvQw4S9IW4KzyOqLVxh2AMSJ6bzIDm/bbgpXfrbXcI5e9o8eVxHSQI5GIiKgtIRIREbXldFZE9EROg70w5UgkIiJqS4hERERtCZGIiKgtIRIREbUlRCIioraESERE1JYQiYiI2hIiERFRW0IkIiJqS4hERERtCZGIiKgtIRIREbW1JkQkLZH0kKStklY2XU9EU7ItxHTSilF8Jc0CPk91N7ch4A5J62w/0K8aMgJptEEbtoV+q7Pt1d3uZvJ23s/12KktRyKnAVttb7P9a2AtsLThmiKakG0hppVWHIkA84DHO14PAac3VMukNJX+MWNN220hXpjaEiLq0uYDZpJWACvKy2ckPdRluVcCP+9hbQejay26vIFKpsF6aUC3Ol7dRCEderktQHvW9cHa73P0exvq8e9rzd9knM81oW2hLSEyBBzf8Xo+sH3kTLZXAavGeiNJg7YHeltePamlu7bU0pY6RujZtgCt/YyTNlM+B8yszwLtuSZyB7BQ0omSDgEuBNY1XFNEE7ItxLTSiiMR2/skfRBYD8wCVtve1HBZEX2XbSGmm1aECIDtG4Ebe/BW4x7i91Fq6a4ttbSljv30cFuAln7GGmbK54CZ9VmQfcA1u4iIiAlpyzWRiIiYhhIiERFRW2uuidQl6feovtE7j6o//XZgne3NjRYW0WfZFqIJ0/qaiKSPAO+hGhpiqDTPp+oWudb2ZX2u5whgCftvxOttP9XnOkQ1fEZnHbe7gT92W9ZJqaU166XX2rYt9IKkuXT8rWzvaLik6GK6h8jDwBtt/2ZE+yHAJtsL+1jLRcClwE3Az0rzfKqB9D5h+5o+1XE28AVgy4g6Xgv8pe2b+lFHqaUV66TU0pr1MhXatC0cLEmnAP8DOIL9/1ZPUf2t7mqqtjratCM1FaZ7iDwInGP70RHtrwZusv36PtbyEHD6yH8Yko4CbrP9uj7VsRk41/YjI9pPBG60/YZ+1FF+ZyvWSfmdrVkvU6FN28LBknQP8O9s3zaifRHwP22f3Exlk9emHampMt2viXwI2ChpC88PWncC1d7lB/tci+gyxhHwW7qPhzRVZvP86YxOPwNe3Mc6oD3rBNq1XqZCm7aFg3XYyAABsP0jSYc1UdBB+E/AW0bbkQISIk2y/X1Jr+P589yi3IPB9nN9LuczwF2SbmL/jfgs4FN9rGM11T0o1nbUcTzVufGr+lgHtGedQLvWS8+1bFs4WN+T9F2q/2A7/1YXAd9vrKp62rQjNSWm9emstil7F+ew/0a83vaePtdxEvCuEXU0cmOjtqyTUktr1kuMTdK5PN/TrPNv1atv8veFpGXAf6Y6nXXAjpTtqxsqrWcSIj3Wph4lko4G3MR/2CPqaM06KfW0Yr3EC0ObdqSmQkKkR0b0KBmi+sfS9x4lkk4A/itwJrC3NB8B/ABYOfLC8hTX0op1UmppzXqJsZXeTB+lOhI5tjTvBG4ALpuOvZratiPVSwmRHmlLjxJJtwKfBa4bPhde7tt9AfAh24v6UUf5va1YJ+V3tma9xNgkracK9zW2nyxt/wx4L7DY9lkNljcpbdqRmioJkR6RtGW0vviSttp+bQvqGHVaA7X0bZ1MoJa+rpcYm6SHRuuSPNa0NmrTjtRUmda9s1qmLT1K7pT0BWDNiDqWAXf3sQ5ozzqBdq2XGNujkj5MdSSyA353Oui97H//+elgJnVX7ipHIj3Uhh4l5RvKy7vVAVxl+9l+1VLqaXydlDpatV5idOVC9Eqqv9VcqusIO6j+Vpfb3t1geZMi6Urgn9N9R+qntqfbd3gOkBCJiFaT9K+pvv9y/3QcnqYtO1JTJSHSI23pUSJpNtUe93nsP1bPDVR73L8ZY/Fe19KKdVJqac16ibFJut32aeX5+4GLgf8FnA387+k4mORMlhDpkbb0KJH0daqeH2vYfzTXZcDRtt/djzpKLa1YJ+X3tma9xNgk3W37zeX5HcDbbe8q1xB+ZPv3m61w4tq0IzVVEiI90pYeJePU8XCfBz1sxTqZQC19XS8xNkn3Am+lumneetsDHdN+FzDTQZt2pKZK7mzYO49K+nDpRQJUPUrKfR762aNkj6QLJP3ubyvpRZLeDfT7G7JtWSfQrvUSYzsCuBMYBI4u/+ki6eVMv/GmFti+fDhAAGw/WU7JndBgXT2TEOmddwPHAH8vaY+k3cAtwNHAn/SxjguB84Edkh4uo7o+CfxxmdZPbVkn8Px6ebKsl4dpbr3EGGwvsP0a2yeWx+H/gH8L/FGTtdXQph2pKZHTWT2k6vak86nO2z7T0b7Edt9HH5V0DNWe22dt/3kDv/904EHbeyW9jKrb5qnAJuC/2N475hv0tpZDqO78tx24CzgX+INSy6pcWI+pMKK78vA1keHuypfNhPGzEiI9IumvqHqRbAZOAS6xfUOZdpftU/tUx7ouzWdSnZfF9rv6UUepZRNwsu19klYBvwS+BSwu7X/cx1q+SvXl2kOpxs46DLi+1CLby/pVSwSApPfZ/nLTdRysfGO9d/4t1c1nnpG0ALhO0gLbn6O/53HnAw8AX6LqxirgXwL/vY81DHuR7X3l+UBHkP6wDAfRT79v+1+Urr4/A15l+zlJfwfc2+daIgA+ASRE4ndmDZ/Csv2IpLdSBcmr6W+IDACXUN1R7T/avkfSP9r++z7WMOzHHXtb90oasD2o6uZJ/T599KJySusw4GVUF293Ay9hZtzZMFpI0n2jTaL6Nv60lxDpnSclnWL7HoByRPJOqjvq9a1fu+3fAldI+mZ53EFzf+f3A5+T9HHg58Ctkh6nuqD4/j7XchXwIDCLKmC/KWkbsAhY2+da4oVjLtW9REZe+xDw//pfTu/lmkiPSJoP7Ovsytcx7Qzb/7eBspD0DuAM2x9r4veXGl4BvIZyn/Om7qUg6VUAtrdLOhJ4G/CY7dubqCdmPklXAV+2/cMu075m+08bKKunEiIREVFbvicSERG1JUQiIqK2hEhERNSWEImIiNoSIhERUdv/B26U2aBzx9jKAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accb4b67908>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df['Amount'].hist(by=df['Class'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Class\n",
      "0    56772\n",
      "1      102\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "print(df.groupby('Class').size())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[<matplotlib.axes._subplots.AxesSubplot object at 0x2accb4bf2860>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4cb6630>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4cda630>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4cfe550>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4d117f0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4d430f0>],\n",
       "       [<matplotlib.axes._subplots.AxesSubplot object at 0x2accb4d65160>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4d785c0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4daf160>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4d6af28>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4debb70>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4e16080>],\n",
       "       [<matplotlib.axes._subplots.AxesSubplot object at 0x2accb4e395c0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4e5c4e0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4e796a0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4e9c710>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4ec1710>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4ee5630>],\n",
       "       [<matplotlib.axes._subplots.AxesSubplot object at 0x2accb4ef88d0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4f25e10>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb6c5e80>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb6e9358>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb70fe80>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb72f940>],\n",
       "       [<matplotlib.axes._subplots.AxesSubplot object at 0x2accbb7539b0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb7799b0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb79d9b0>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accb4d2b358>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb7de748>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb7f11d0>],\n",
       "       [<matplotlib.axes._subplots.AxesSubplot object at 0x2accbb8266d8>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb84b6d8>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb865d68>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb88f278>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb8b37b8>,\n",
       "        <matplotlib.axes._subplots.AxesSubplot object at 0x2accbb8d5cf8>]], dtype=object)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA30AAANeCAYAAACmsmchAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4wLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvpW3flQAAIABJREFUeJzs3X+cVdV97//XW1BjIghqIQRsxtyQFpX4A6rk1lqIiaJJQ3rRBEMFIpam1SZeuU2w6f1qkxiIt8ZgYsxXoxGsEa1prhRQYtS5bbj+xl9RYhh1GkbxVwQcIqjA5/6x19HDcGbmzMyZc/Y5834+Hucx56y99pm1zlmzZ6+1P2ttRQRmZmZmZmbWmPaqdQHMzMzMzMys/7jTZ2ZmZmZm1sDc6TMzMzMzM2tg7vSZmZmZmZk1MHf6zMzMzMzMGpg7fWZmZmZmZg3MnT6zOiXpIkn/XOtyWH2T9PeSfljrcpiZmVn/caevAiQ1S9okad9al6UjSXMk/aLW5bDek/Q5SQ9K2ippo6TbJB1f63JZfUjtpvDYJWlb0euZEfHNiDi71uW0+iRptaSvlUifJukFSR+XdLekLZJaa1BEy5m+thlJTWn765J+JeljVSm45UYF2tDXJT0uaYeki6pR5jxwp6+PJDUBfwIE8KmaFsYajqTzge8A3wRGAr8PfB+YVstyWf2IiP0LD+A3wJ8Vpd1Q6/JZ3bsOOFOSOqSfCdwAbAGuBf6uyuWy/LqOvrWZG4GHgYOArwK3SPq9/imq5dR19K0NtQBfBlb2VwHzyJ2+vpsF3EvWAGcXEiVdJ+n76arMVklrJL1X0nfSVcFfSTq6KP+4dMVws6QnJH2qaFuzpLOLXu929U5SSPqCpPXpva9QZhzwA+AjqQyb+/ejsEqSdADwNeCciPjXiPhdRLwVEf8WEXscyCT9Sxrh2iLp3yUdXrTtVElPSmqX9Jyk/5HSD5a0IrW7VyX9hyQfFwaQ4jDhNIIekj4vaUM6nnxB0h9Jeiy1k+912P8sSetS3tWS3l+bmliN/G/gQLLBTwAkDQc+CSyNiPsj4nrgmRqVz/Kn121G0oeAY4ALI2JbRPwEeByYXpWSW1706bgTEUsi4jagvRqFzQuf3PXdLLJRhRuAkyWNLNr2GeAfgIOBN4B7gLXp9S3AtwEk7Q38G/AzYATwt8ANkv6gB+X4JPBHwJHp954cEeuALwD3pFH9Yb2tpNXER4B3AT8tM/9twFiyNrSWrE0WXAP8VUQMAY4A7krp84E24PfIriT+PdlVaxvYjiNrS58lu9L8VeBjwOHAZyT9KYCkT5O1mf9G1ob+g2wU3gaIiNgG3Ez2v7DgM8CvIuLR2pTK8qyPbeZw4JmIKD5ZfzSl2wDh407vuNPXB2le1fuBmyPiIeBp4HNFWX4aEQ9FxHayE/ftEbE0InYCNwGFK32TgP2BRRHxZkTcBawAzuhBcRZFxOaI+A1wN3BUnypneXAQ8EpE7Cgnc0RcGxHtEfEGcBFwZLpaCPAWcJikoRGxKSLWFqWPAt6friL+R0S402dfj4jtEfEz4HfAjRHxUkQ8R9axKxy7/gpYGBHrUjv9JnCUr/YNOEuA0yXtl17PSmlmneltm9mfLHSv2BZgSAXLZvXBx50ecqevb2YDP4uIV9LrH1MU4gm8WPR8W4nX+6fn7wM2RMSuou3/CYzuQVleKHr+etF7W/36LXCwpMHdZZQ0SNIiSU9Leg1oTZsOTj+nA6cC/ynp/0j6SEr/X2Sx7T+T9IykBZWtgtWpco9d7wcWp7DPzcCrgOjZscvqXET8AngZmCbpA2RRJz+ubaksz/rQZrYCQzukDWWAhemZjzu90e3JpJWWRhY+AwySVOhw7QsMk3RkD9/ueeAQSXsVdfx+H/h1ev474N1F+d/bg/f2VZv6dQ+wHfg0WThwVz5HtrjLx8g6fAcAm8hOwImIB8gOjHsD55KFRRySQmTmA/PTHMC7JT0QEXdWvjrWgDYAF3tBGAOWko20/wHZYOiL3eQ3602beQL4gKQhRSGeR+KT/YHKx50e8JW+3vs0sBM4jCyU8ihgHFno06wu9ivlPrKO3Zcl7S1pMvBnwLK0/RHgv0l6t6QPAnN78N4vAmMk7dPDMlmNRcQW4P8DrpD06fT97y3pFEmXdMg+hGze6G/JBgi+WdggaR9JMyUdEBFvAa+RtV0kfVLSB9MKWIX0nf1fO2sQPwAuKCwaJOkASafXuExWG0vJBp3+kqIQK0l7SXoXsHf2Uu/y/yNLetxmIuLXZOdEF6b0Pwc+DPyk6qW3POjVcSedS72LrB80OG0fVOWyV507fb03G/hRRPwmIl4oPIDvATPpwVXUiHiT7HYPpwCvkC3JPysifpWyXAa8SdaBW8LuC3R05y6ykbEXJL3SXWbLl4j4NnA+2YJAL5NdWTmXbOWqYkvJQoKfA54kW1G22JlAawr9/ALwFyl9LPBzspCZe4DvR0RzxStiDSkifgp8C1iW2tYvyY5jNsBERCvwf4H3AMuLNp1AFhK8iiyCZRvZomU2wPWhzcwAJpJFsywCTouIl6tQZMuZPrShq1PaGWQLlW0jO09qaPKaDWZmZmZmZo3LV/rMzMzMzMwamDt9ZmZmZmb9LM0du1/So5KekPSPKf1QSfdJWi/ppsL8M0n7ptctaXtT0XtdkNKfknRyUfrUlNbiFbmtmDt9ZmZmZmb97w3goxFxJNkCgFMlTSKbG31ZRIwlm6tYWLBvLrApIj5Itr7DtwAkHUY2t/FwYCrw/XTrpkHAFWRzqw8Dzkh5zdzpMzMzMzPrb5HZml7unR4BfJR3bs20hGyFeMhuxVRYlfIW4MS02vY0YFlEvBERz5Ldb/fY9GiJiGfSIoHLUl6z+r1P38EHHxxNTU27pf3ud7/jPe95T20KlAP1Uv+HHnrolYj4vVqXo5HbUKPUA0rXxW0oH+q5rnluQx3V6+fc6OWupzbUmXr5jhqpnBHBunXr2Guvvdi1a9c24Glgc0TsSFnagNHp+WiyVbuJiB2StgAHpfTiVbqL99nQIf24UuWQNA+YB7DffvtNOOSQQ8qrZD/btWsXe+01sK5J9bXOv/71r8s6FtVtp6+pqYkHH3xwt7Tm5mYmT55cmwLlQL3UX9J/1roM0NhtqFHqAaXr4jaUD/Vc1zy3oY7q9XNu1HLv3LmTiRMnQnaVBkmHkl1RORBYC5wZEW9K2pfsdjoTyO6h+tm0xDySLiAL3dsJfDEiVqf0qcBiYBDww4hY1F15y2lDva1rXjRiOTdv3szw4cN3kN3juaPC0vrqZFtn6aV6DiWX6Y+Iq4CrACZOnBi9bUOVVi/fdSX1tc7l/j8bWF1pMzMzsz5YvHgx48btdp7u+VjWY8OGDQNoByYBwyQVLsSMAZ5Pz9uAQwDS9gOAV4vTO+zTWbqZO31mZmZm5Whra2PlypWcffbZAKT5VZ6PZWV5+eWX2bx5MwDbtm0DGAqsA+4GTkvZZgO3pufL02vS9rsiu8H2cmBGWt3zUGAscD/wADA2rQa6D9ngQvFNy20Aq9vwTjMzM7NqOu+887jkkktob28vJB1EjedjjRw5kubm5l7VZ+vWrb3et5oapZxPP/00ixYtYteuXezatQvgtYhYIelJYJmkbwAPA9ekXa4BrpfUQnaFbwZARDwh6WbgSWAHcE5E7ASQdC6wmixE+NqIeKIfqmp1yJ0+MzMzs26sWLGCESNGMGHChOIT+87mVnW1reLzsXo7H6he5k81SjknT57M3Llz334taSNARDxDdqV3NxGxHTi91HtFxMXAxSXSVwGrelh0GwAaqtP3+HNbmLNgZbf5Whd9ogqlsXrkNlR/JP134Gyyk6PHgc8Do6jQwgpWfU1l/A2C/w6t70q1tfnjd+zxf6B10SdYs2YNy5cvZ9WqVWzfvh1gCPAd0nysdLWv1HystjLnY9FF+oDVtGBlye+kIx8P8q3c4/p1U/O/Smu98pw+M6tbkkYDXwQmRsQRZOEsM6jQwgrVrIuZ5dvChQtpa2ujtbWVZcuWAbRHxEw8H8vM6oA7fWZW7wYD+6WR9HcDG6ncwgpmZt35CnB+mnd1ELvPxzoopZ8PLIBsPhZQmI91O2k+VrpSWJiPtQ642fOxzKxSGiq808wGloh4TtI/Ab8BtgE/Ax6isgsr7Ka7BRTyvuDA489tKTvv+NEHdLm9v+o6f/yO7jNBrj9na2xp3lYLeD6WmdUHd/rMrG5JGk52le5QYDPwL2T3uOqotwsr7JnYzQIKeV9woJw5qwWtMyd3ub2/6lpuGbsrn5mZmWUc3mlm9exjwLMR8XJEvAX8K/BfqdyNbs3MzMzqnq/0mVk9+w0wSdK7ycI7TwQe5J2FFZZRemGFeyhaWEHScuDHkr4NvI93FlaoG+WujFZL9VBGMzOzRuROn5nVrYi4T9ItZLdl2EF2U9urgJVU6Ea3teaOkpmZmfWVO31mVtci4kLgwg7JFVtYwczMzKzeeU6fmZmZmZlZA3Onz8zMzMzMrIG502dmZmZmZtbA3OkzMzMzMzNrYO70mZmZmZmZNTB3+szMzMzMzBqYO31mZmZmZmYNzJ0+MzMzMzOzBuZOn5mZmZmZWQNzp8/MzMzMzKyBDa51AczMzMzMzB5/bgtzFqwsK2/rok/0c2kaizt9ZmZWl5p6cWKwc+dOJk6cCPBBAEmHAsuAA4G1wJkR8aakfYGlwATgt8BnI6I17XMBMBfYCXwxIlan9KnAYmAQ8MOIWNTnSpqZmVWAwzvNzGzAWLx4MePGjStO+hZwWUSMBTaRdeZIPzdFxAeBy1I+JB0GzAAOB6YC35c0SNIg4ArgFOAw4IyU1xrE9u3bOfbYYznyyCM5/PDDAd4H2cCBpPskrZd0k6R9Uvq+6XVL2t5UeC9JF6T0pySdXJQ+NaW1SFpQ3RqaWSNzp8/MzAaEtrY2Vq5cydlnnw2AJAEfBW5JWZYAn07Pp6XXpO0npvzTgGUR8UZEPAu0AMemR0tEPBMRb5JdPZxWhWpZley7777cddddPProozzyyCMAQyVNwgMHZlYHHN5pZmYDwnnnnccll1xCe3t7IekgYHNE7Eiv24DR6floYANAROyQtCXlHw3cW/S2xfts6JB+XKlySJoHzAMYOXIkzc3NXZZ769at3ebJo3oo9/zxO/ZIG7nfnukd67F9+3YAAUE2cPC5tGkJcBFwJVmn/6KUfgvwvY4DB8CzkgoDB5AGDgAkFQYOnux1Bc3MEnf6zMys4a1YsYIRI0YwYcKE4hN4lcga3WzrLL1U5EyUSCMirgKuApg4cWJMnjy584KTdTi6y5NH9VDuUgtGzB+/g0sf3/30qHXmZCCbEzphwgRaWloAXgOepg4GDjpTLx3zUh3xjvJQj3r4PG3gcqfPzOqapGHAD4EjyE6yzwKeAm4CmoBW4DMRsSmNsi8GTgVeB+ZExNr0PrOBf0hv+42IWII1jDVr1rB8+XJWrVpVuEozBPgOMEzS4HTSPgZ4Pu3SBhwCtEkaDBwAvFqUXlC8T2fp1iAGDRrEI488wubNmxk+fPh7gHElsuVu4KAz9dIxL9UR76jQMa+levg8beAqe05fijd/WNKK9NoTl80sDxYDt0fEHwJHAuuABcCdaY7Nnek1ZHNlxqbHPLIQLCQdCFxINqp+LHChpOHVrIT1r4ULF9LW1kZrayvLli0DaI+ImcDdwGkp22zg1vR8eXpN2n5XRERKn5H+1x1K1pbuBx4Axqb/jfuQzdlaXo26WfUNGzYMoB2YRBo4SJtKDRxQ5sBBVwMKZmZ90pOFXL5EdjJV4InL1iM7d+7k6KOPhqKl0j1wYH0haShwAnANQES8GRGb2X0Rjo6LcyyNzL1kJ2ujgJOBOyLi1YjYBNxBdpyyxvcV4Pw0r+ogUltKPw9K6eeTBg4i4gngZrJ5VrcD50TEznSl8FxgNdn/yptTXmsQL7/8Mps3bwZg27ZtAEPJvmsPHFhZNmzYwJQpUxg3blxhBdgRkA08SrojnQ/dURh0VObydG7zmKRjCu8laXbKvz5FqhTSJ0h6PO1zeYpwMSsvvFPSGOATwMVk/xwLK5554rKVrbBUelr1DN4ZOFgm6QdkAwZXUjRwIGlGyvfZDgMH7wN+LulD6b2uAD5ONlL6gKTlEeE21Pg+ALwM/EjSkcBDZANUIyNiI0BEbJQ0IuV/e45NUphL01n6HrqbS1PpOR3dzWPpT5VeYKRWdemkjC0A6X/PsR03RsR24PRSO0bExWT/DzumrwJW9aGolmMbN25k9uzZ7Ny5k127dgG8FhErJD0JLJP0DeBhdh84uD6d77xK9v+LiHhCUmHgYAdp4ABAUmHgYBBwrQcOGsvgwYO59NJLOeaYY2hvb2fo0KEj0rnNHLLolEVp4HoB2YBUcXTKcWTnSMcVRadMJAsBfiid92xKeeaRzRtdRTaAeVtVK2q5VO6cvu8AXyabAwE5XfGsnIm+kI/Jvv0hzxOIX375Zf75n/+Zv/iLv+DGG2/EAwdWIYOBY4C/jYj7JC3mnVDOUno6x2bPxG7m0lR6TkephSaqpbs5Mj2ta63qkoe5Plb/PvzhD/Pwww+//VpSYWDJAwdWllGjRjFq1CgAhgwZArCN7Fx4GjA5ZVsCNJN1+t6OTgHulVSITplMik4BkHQHMFVSMzA0Iu5J6UvJIl3c6bPuO32SPgm8FBEPSZpcSC6RteYTl797w63dTvSFxj0ByPME4tNOO42rr74690ulN8rAQZ4HAHqqm7q0AW0RcV96fQtZp+9FSaPSVb5RwEtF+TubSzO5Q3qnv9TMzKwvWltbAd4N3EflolNGp+cd0/fQ3flQpZUb6VHueRjk/1ysXNU6ZyvnSt8fA5+SdCrwLrIYdq94ZmWrp6XSG2XgIM8DAD3VVV0i4gVJGyT9QUQ8BZxIdoX3SbK5NIvYc47Nuelq8HHAlvQPdjXwzaLFW04CLuivOpmZ2cC1detWpk+fDrAhIl7rYtpdT8+HKha1UmnlRnqUs1JrQd7PxcpVrXO2bhdyiYgLImJMRDSRxaPf5RXPrCcKS6U3NTUxY8YM6LBUesrmFc+st/4WuEHSY8BRwDfJOnsfl7SebK7nopR3FfAM2Xyuq4G/AUghMl8nOx49AHytEDZjZmZWKW+99RbTp09n5syZAJtT8ospKoUeRKd0lj6mRLpZj1bv7MgrnllZvFS69aeIeCQiJkbEhyPi0xGxKSJ+GxEnRsTY9PPVlDci4pyI+C8RMT4iHix6n2sj4oPp8aPa1cjMzBpRRDB37lzGjRvH+eefX7yp+Lyn4/nQrLSK5yRSdArZOfNJkoanCJWTgNVpW7ukSWkthFlF72UDXI9uzh4RzaR5Lp64bBXwFbzimZmZmQ0Aa9as4frrr2f8+PEcddRRAIel6VOLgJslzQV+wzvn0auAU8miU14HPg9ZdIqkQnQK7B6d8tfAdcB+ZAu4eBEXA3rY6TPrqxSz7KXSzczMbEA5/vjjyQKXMpKeTOcvkM1J302Kcjqn1HtFxLXAtSXSHwSOqEiBraH0JbzTzMzMzMzMcs6dPjMzMzMzswbmTp+ZmZmZmVkDc6fPzMzMzMysgbnTZ2ZmZmZm1sDc6TMzMzMzM2tg7vSZmZmZmZk1MHf6zMzMzMzMGpg7fWZmZmZmZg3MnT4zMzMzM7MG5k6fmZmZmZlZA3Onz8zMzKwbGzZsYMqUKYwbN47DDz8cYASApAMl3SFpffo5PKVL0uWSWiQ9JumYwntJmp3yr5c0uyh9gqTH0z6XS1K162lmjcmdPjMzM7NuDB48mEsvvZR169Zx7733AoyQdBiwALgzIsYCd6bXAKcAY9NjHnAlZJ1E4ELgOOBY4MJCRzHlmVe039Rq1M3MGp87fWZmZmbdGDVqFMcck12sGzJkCMA2YDQwDViSsi0BPp2eTwOWRuZeYJikUcDJwB0R8WpEbALuAKambUMj4p6ICGBp0XuZmfWJO31mVvckDZL0sKQV6fWhku5LoVM3Sdonpe+bXrek7U1F73FBSn9K0sm1qYmZ1YPW1laAdwP3ASMjYiNA+jkiZRsNbCjarS2ldZXeViLdzKzPBte6AGZmFfAlYB0wNL3+FnBZRCyT9ANgLlnY1FxgU0R8UNKMlO+zKURrBnA48D7g55I+FBE7q10RM8u3rVu3Mn36dIANEfFaF9PuSm2IXqTv+cbSPLIwUEaOHElzc3M3pS5t69atvd63WuaP38HI/bKfXclDPerh87SBy50+M6trksYAnwAuBs5PCx98FPhcyrIEuIis0zctPQe4Bfheyj8NWBYRbwDPSmohm2tzT5WqYWZ14K233mL69OnMnDmTtWvXbk7JL0oaFREbU4jmSym9DTikaPcxwPMpfXKH9OaUPqZE/j1ExFXAVQATJ06MyZMnl8rWrebmZnq7b7XMWbCS+eN3cOnjXZ+yts6cXJ0CdaEePk8buNzpM7N69x3gy8CQ9PogYHNEFIaFi0Ok3g6riogdkrak/KOBe4ves9Owqu5G2Cs90tvd6HZ/6q4ePa1rrerikXerhIhg7ty5jBs3jvPPP5/58+cXNi0HZgOL0s9bi9LPlbSMbNGWLaljuBr4ZtHiLScBF0TEq5LaJU0iCxudBXy3OrUzs0bnTp+Z1S1JnwReioiHJE0uJJfIGt1sKzusqrsR9kqP9M5ZsLJi79VT3Y2c97SutapLHq4AWP1bs2YN119/PePHj+eoo44COEzSqWSdvZslzQV+A5yedlkFnAq0AK8DnwdInbuvAw+kfF+LiFfT878GrgP2A25LDzOzPnOnz8zq2R8Dn0onXu8im9P3HbJV8ganq33FIVKFcKs2SYOBA4BX6TwMy8wMgOOPP55sUc2MpCcjYlV6eWLH/GkFznNKvVdEXAtcWyL9QeCIihTYzKyIV+80s7oVERdExJiIaCJbiOWuiJgJ3A2clrJ1DLcq3Aj5tJQ/UvqMtLrnoWT3x7q/StUwMzMz61e+0mdmjegrwDJJ3wAeBq5J6dcA16eFWl4l6ygSEU9Iuhl4EtgBnOOVO83MzKxRuNNnZg0hIprJVsAjIp4hW32zY57tvDPfpuO2i8lWADUzMzNrKO70mZmZmVlVNdVwkSqzgchz+szMrOFt2LCBKVOmMG7cOA4//HCAEQCSDpR0h6T16efwlC5Jl0tqkfSYpGMK7yVpdsq/XtLsovQJkh5P+1yuLu7abWZmVk2+0mdmZn1SDyP2gwcP5tJLL+WYY46hvb2doUOHjpB0GDAHuDMiFklaACwgmxN6CtmCPmPJ7rF2JXCcpAOBC4GJZLf1eEjS8ojYlPLMI7vn4ypgKl5y38zMcsCdPjOzGqiHjlIjGTVqFKNGjQJgyJAhANuA0cA0YHLKtoRsXuhXUvrStLrrvZKGSRqV8t5RuK+apDuAqZKagaERcU9KXwp8Gnf6zMwsB7rt9Ek6BFgKvBfYBVwVEYvTaOdNQBPQCnwmIjalcJbFZDckfR2YExFr03vNBv4hvfU3ImJJSp/AOzcjXQV8KYpvhmN1bcOGDcyaNYsXXniBvfbaC4rCqnAbMrMqa21tBXg3cB8wMiI2AkTERkkjUrbRwIai3dpSWlfpbSXS9yBpHtkVQUaOHElzc3OX5d26dWu3efKoHso9f/yOPdJG7rdnet7rYWbWnXKu9O0A5kfEWklDyEJZ7sAhMVYmh1WZWV5s3bqV6dOnA2yIiNe6mHZXakP0In3PxIirgKsAJk6cGJMnT+6yzM3NzXSXJ4/qodxzSlxxnz9+B5c+vvvpUevMyVUqkTW6s846ixUrVjBixIi30zwIbtXQ7UIuEbGx0MAioh1YxzshMUtStiVkYSxQFBITEfcChZCYk0khMekkvRASM4oUEpMa5dKi97IGMGrUKI45JlsDoURYlduQmVXFW2+9xfTp05k5cybA5pT8YjqGkH6+lNLbgEOKdh8DPN9N+pgS6WZmb5szZw633357x+QFZIPgY4E702vYfRB8HtkAN0WD4MeR3Z7owsIiVLwzCF7Yb2q/VcbqSo/m9ElqAo4mpyExpUIySmnUMI16CKV54YUXoIZhVWY2MEUEc+fOZdy4cZx//vnMnz+/sGk5MBtYlH7eWpR+rqRlZCdWW9JxajXwzaITrJOACyLiVUntkiaRHd9mAd+tTu3MrF6ccMIJhRDzYp5bbP2u7E6fpP2BnwDn5TUk5rs33LpHSEYpjRqmkfdQmq1bt/Knf/qnUMOwqoEycFAPAwDlaqS6WO2sWbOG66+/nvHjx3PUUUcBHCbpVLLO3s2S5gK/AU5Pu6wiC6lqIQur+jxA6tx9HXgg5fta4cQL+GveCau6DZ9omVl5cnchpdLKObeC8s/DIP/nYuWq1nlOWZ0+SXuTdfhuiIh/TckvShqVGme5ITGTO6Q345CYAaE4rGrt2rW7hVVVsw0NlIGDvA8A9EQj1cVq5/jjj6d4WoukJyNiVXp5Ysf8aWT9nFLvFRHXAteWSH8QOKIiBTYzy9Hc4r4qNX+2lFJzajuT93OxclXrPKfbOX1pEuk1wLqI+HbRpkJIDOwZEjMr3dh2EikkBlgNnCRpeAqLOQlYnba1S5qUftesoveyBtAxrKqI25CZmZkNdJ5bbP2u204f8MfAmcBHJT2SHoWQmI9LWg98PL2GLCTmGbKQmKuBv4EsJAYohMQ8wJ4hMT9M+zyNQ2IaSiGs6q677ioVVuU2ZGZmZgOZB8Gt33V7/TQifkHpy8XgkBgrg8OqzMzMzOCMM86gubmZV155BeDDaT6x5xZbv+vR6p1mZmZmZtY7N95449vPJT0WEdeklx4Et35VTninmZmZmZmZ1Sl3+szMzMzKcNZZZzFixAiOOOKdCymSDpR0h6T16efwlC5Jl0tqkfSYpGOK9pmd8q+XNLsofYKkx9M+l6uLexuZmfWEO31mVrckHSLpbknrJD0h6UspvWInYWZmBXPmzOH222/vmLwAuDMixgJ3ptcApwBj02MecCVkxyfgQuA44FjgwsIxKuWZV7Tf1H6rjJkNKO70mVk92wHMj4hxwCTgHEmHUdmTMDMzAE444QQOPPDAjsnTgCXp+RLg00XpSyNzLzAsLceC1+gCAAAgAElEQVR/MnBHRLwaEZuAO4CpadvQiLgnzeVaWvReZmZ94oVczKxupeWpN6bn7ZLWAaPJTrYmp2xLgGbgKxSdhAH3SiqchE0mnYQBSLqDbIT9nRn3ZXr8uS1l34TWzBrCyHQsIiI2ShqR0kcDG4rytaW0rtLbSqTvQdI8soErRo4cSXNzc68KvnXr1l7v21fzx+8oO+/I/brPX6t6FKvl52nWHXf6zKwhSGoCjgbuo3InYaV+T5cnW+WcnNSL797Q9e2dRu6X5Zk/vkoF6iWfhFmNlJqPF71I3zMx4irgKoCJEyfG5MmTe1XA5uZmertvX/VkcGz++B1c+njXp6ytMyf3sUR9V8vP06w77vSZWd2TtD/wE+C8iHiti7UP+v1k67s33NrtyUmjKOdELA/ycDJoDe1FSaPSANMo4KWU3gYcUpRvDPB8Sp/cIb05pY8pkd/MrM88p8/M6pqkvck6fDdExL+m5BfTyRc9OAkrlW5m1p3lQGHxp9nArUXps9ICUpOALSkCYTVwkqThae7wScDqtK1d0qS0auesovcyM+sTd/rMrG6lE6NrgHUR8e2iTRU5CatKJcysbpxxxhl85CMf4amnngL4sKS5wCLg45LWAx9PrwFWAc8ALcDVwN8ApLnDXwceSI+vFeYTA38N/DDt8zRwWzXqZWaNL/9xOWZmnftj4EzgcUmPpLS/JzvpujmdkP0GOD1tWwWcSnZC9TrwechOwiQVTsJg95MwMzMAbrzxnbWdJD0WEdeklyd2zJsWjDqn1PtExLXAtSXSHwSO2HMPM7O+cafPzOpWRPyC0vPxoEInYWZmZmb1zuGdZmZmZmZmDcydPjMzMzMzswbmTp+ZmZmZmVkDc6fPzMzMzMysgbnTZ2ZmZmZm1sDc6TMzMzMzM2tg7vSZmZmZmZk1MN+nz8zMzMzMeqxpwcpaF8HK5Ct9ZmZmZmZmDcydPjMzMzMzswbm8E4zMzNrKA45MzPbna/0mZmZmZmZNTB3+szMzMzMzBqYwzvNzMzMzKyulBvG3broE/1ckvrgK31mZmZmZmYNzJ0+MzMzMzOzBpabTp+kqZKektQiaUGty2P1x23I+sptyPrKbcj6ym3I+sptyErJRadP0iDgCuAU4DDgDEmH1bZUVk/chqyv3Iasr9yGrK/chqyv3IasM3lZyOVYoCUingGQtAyYBjzZH7/MEz8bUlXbkDUktyHrK7ehfjYA7r/nNtRLPrd7m9uQlZSXTt9oYEPR6zbguI6ZJM0D5qWXWyU91SHLwcArlSqUvlWpd6qaita/H72/H96zqm2oDtpGvbSFcpSqS923oUbwxTqpayd/r3luQx3VxedcQl2Wu1S7boA21Jm6+I4qeazp5//fPS1nI7ShiuqP/ysD4JytrHaUl06fSqTFHgkRVwFXdfom0oMRMbGSBasnA7z+bkNFGqUeUNW6uA310ECqa5kq0ob2eNM6/Zxd7t79+hJpfW5Dnf6yOvmOXM6eFaNEWr+1oUrLyWdYVdWqcy7m9JGNQhxS9HoM8HyNymL1yW3I+sptyPrKbcj6ym3I+sptyErKS6fvAWCspEMl7QPMAJbXuExWX9yGrK/chqyv3Iasr9yGrK/chqykXIR3RsQOSecCq4FBwLUR8UQv3ip3l6mrbMDW321oD41SD6hSXdyGemUg1bVbFWxDHdXr5+xy91A/tqHO1Mt35HKWqQZtqNJq/hnWQFXqrIg9wnzNzMzMzMysQeQlvNPMzMzMzMz6gTt9ZmZmZmZmDawhOn2Spkp6SlKLpAW1Lk8lSWqV9LikRyQ9mNIOlHSHpPXp5/CULkmXp8/hMUnHFL3P7JR/vaTZtapPHnTXXiTtK+mmtP0+SU3VL2V5yqjLHEkvp/bziKSza1HO7ki6VtJLkn7ZyfZO23Ze1NNxqNTnXcnjiqQJ6bjVkvZVV7/Duibpf0gKSQen17n+e5D0vyT9KpXtp5KGFW27IJX7KUkn17KcHdXT33Bf1NP3k9fvRNIhku6WtE7SE5K+lNJ9jOulvH7XldaT/78VFxF1/SCbpPo08AFgH+BR4LBal6uC9WsFDu6QdgmwID1fAHwrPT8VuI3sHi2TgPtS+oHAM+nn8PR8eK3rltf2AvwN8IP0fAZwU63L3Ye6zAG+V+uyllGXE4BjgF92sr1k287Lo96OQ6U+70oeV4D7gY+kfW4DTunqd/jR5Xd1CNmCDP9Z+F9QB38PJwGD0/NvFbWlw9Lfxr7AoelvZlCty5vKVld/wwPh+8nzdwKMAo5Jz4cAv06fn49xDfZd90Ndy/7/W+lHI1zpOxZoiYhnIuJNYBkwrcZl6m/TgCXp+RLg00XpSyNzLzBM0ijgZOCOiHg1IjYBdwBTq13onCinvRR/vrcAJxauVORMw7T9iPh34NUusnTWtvOirr6LTj7vihxX0rahEXFPZP/BlnZ4r1K/wzp3GfBldr+5cq7/HiLiZxGxI728l+w+YZCVe1lEvBERzwItZH87eVBXf8N9UUffT26/k4jYGBFr0/N2YB0wGh/jeiu333Wl9fD/b0U1QqdvNLCh6HVbSmsUAfxM0kOS5qW0kRGxEbIDDzAipXf2WTT6Z9QT5XwWb+dJ/xi3AAdVpXQ9U+73Oj2F8dwi6ZAS2+tB3ttw3stXjkodV0an5x3Tu/odVoKkTwHPRcSjHTbVU3s7i+yqJOS73HkuW3/K8/eTt/KUlKaAHA3ch49xvVUX33U/qkq7ycV9+vqo1BWYRroPxR9HxPOSRgB3SPpVF3k7+ywa/TPqiXI+i3r5vMop578BN0bEG5K+QDaC9NF+L1nl5f07yXv5+qKnx5VG/iwqTtLPgfeW2PRV4O/JQvH22K1EWlU/467KHRG3pjxfBXYANxR2K5E/L20jz2XrsQb5fvJWnj1I2h/4CXBeRLyWz6CgupD777oRNEKnr41szkPBGOD5GpWl4iLi+fTzJUk/JbsE/qKkURGxMYX0vJSyd/ZZtAGTO6Q393PR86qc9lLI0yZpMHAAXYce1kq3dYmI3xa9vJps/kY9yvvfed7LV45KHVfaeCdcrDh/V79jwIqIj5VKlzSebF7Vo+lEcgywVtKx5KC9dVbugrSwzyeBE1OYL+Sg3F3Ic9l6rEG+n7yVZzeS9ibr8N0QEf+akn2M651cf9dVUJV20wjhnQ8AYyUdKmkfsoU3lte4TBUh6T2ShhSek434/pKsfoWV8mYDt6bny4FZaWW3ScCWdJl4NXCSpOFpRaCTUtpAVE57Kf58TwPuKvqnmCfd1qXDPJ9Pkc07qEedte28aITjUEWOK2lbu6RJaS7srA7vVep3WAcR8XhEjIiIpohoIjspOiYiXiDnfw+SpgJfAT4VEa8XbVoOzFC2QvKhwFiyRX/yoBH+hstSR99Pbr+TdGy7BlgXEd8u2uRjXO/k9ruukuq0m/5YHabaD7KVzH5NtvLPV2tdngrW6wNkKxg9CjxRqBvZ/LI7gfXp54EpXcAV6XN4HJhY9F5nkU3KbgE+X+u65a29AF8j+wcI8C7gX9JndT/wgVqXuQ91WZjazqPA3cAf1rrMndTjRmAj8BbZye1c4AvAF9L2Ttt2Xh71dBzq5POu2HEFmEg2QPU08D1AKb3k7/CjrO+slXdW78z130NqDxuAR9LjB0XbvprK/RRpVde8POrpb3igfD95/U6A48nCDx8r+hxP9TGu8b7rfqhn2f9/K/0o/CM2MzMzMzOzBtQI4Z1mZmZmZmbWCXf6zMzMzMzMGpg7fVUiabWkr5VInybpBUkfl3S3pC2SWjvkGSHpRknPp+1rJB1XtcJbLvSlDaV8d0t6WdJrkh6V1JA3PrXO9bUNFeX/U0kh6Rv9WmDLnQoch1olbZO0NT1+VpWCW25U4jgk6UuSnpX0O0nrJH2o3wtuudHHc+rfLzr+FB4haX7VKlAj7vRVz3XAmWnFp2Jnkt0jZwtwLfB3Jfbdn2xlownAgWT3Wlup7P4wNnBcR+/bEMCXgFERMRSYB/xzh9U9rfFdR9/aUGGZ8sVkNyK2gec6+tiGgD+LiP3To9R9CK2xXUcf2pCks8kWv/gE2fnRJ4FX+quwlkvX0cs2FBG/KTr+7A+MB3aR3X6jobnTVz3/m6zD9ieFhLTM+SeBpRFxf0RcDzzTcceIeCYivh0RGyNiZ0RcBewD/EGVym750Os2BBARj0XEjsJLYG92vy+ONb4+taFkPvAz4Ff9WVDLrUq0IRvYet2GJO0FXAj894h4MjJPR0Qe76Vr/aeSx6FZwL9HRGt/FDRP3OmrkojYBtxM1rgKPgP8KiIe7cl7STqKrNPXUrkSWt5Vog1JWiFpO9lVmmbgwUqX0/Krr21I0vvJbtOwR1iNDQwV+l92Qwo1/5mkIyteSMu1PrahMelxhKQNKcTzH1Nn0AaISp5Tp/dYUqmy5Zn/SKprCXC6pP3S6x43NElDgeuBf4yILRUun+Vfn9pQRHwSGEJ2P5zVEbGr8kW0nOtLG7oc+J8RsbVfSmb1oi9taCbQBLyf7N6hqyUNq3gJLe9624bGpJ8nkYXlTQHOIAv3tIGlEufUfwKMBG6pcNlyyZ2+KoqIXwAvA9MkfQD4I+DH5e6fGva/AfdGxML+KaXlWV/bUHqPtyLiNuBkSZ/qh2JajvW2DUn6M2BIRNzUz0W0nOvLcSgi1kTEtoh4Pf0f20xRiJYNDH1oQ9vSz0siYnMKyfv/yQYybQCpxPkQMBv4yUAZyBxc6wIMQEvJRiP+APhZRLxYzk6S9iWLYX4O+Kv+K57VgV61oRIGA/+lYqWyetKbNnQiMFHSC+n1AcBOSeMjwivBDjyVOg4F0HExBhsYetOGngLeJGs3Zr0+DqULKacDf95PZcsdX+mrvqXAx4C/pOgytKS9JL2LbHENSXqXpH3Str3JLj1vA2Y5JG/A600b+kNJp0jaT9Lekv4COAH4PzUov9Vej9sQ8D+BDwFHpcdy4Grg89UsuOVGb45Dvy/pjyXtk9L/DjgYWFOD8lvt9bgNRcTrwE3AlyUNkTQm7b+i6qW3POjN/7KCPyeLNLi7WoWtNXf6qiyFIvxf4D1kJ00FJ5B16lYBv5+eF+5f9F/JViQ6CdhcdF8Rh8QMQL1sQwIuAl4iC4f4EvDZiFhblUJbrvSmDUVEe0S8UHikbb/zqnkDUy+PQ0OAK4FNZFErU4FTIuK31Sm15Ukv2xDAucBW4HngHrKQvmv7v8SWN31oQ5CFdi6NiAFz1VgDqK5mZmZmZmYDjq/0mZmZmZmZNTB3+szMzMzMzBqYO31mVjckDZL0sKQV6fWhku6TtF7STUULRuybXrek7U1F73FBSn9K0slF6VNTWoukBdWum5mZmVl/cafPzOrJl4B1Ra+/BVwWEWPJFoco3KB3LrApIj4IXJbyIekwYAZwONkiEt9PHclBwBXAKcBhwBkpr5mZmVndq9v79B188MHR1NRU62KU9Lvf/Y73vOc9tS5Gr1Sj7A899NArEfF7/fpLytDfbSgP7aDWZajk73/zzTcZMmQI733ve1m/fv2bkgR8FPhcyrKEbIXSK4Fp6Tlktzv5Xso/DVgWEW8Az0pqAY5N+Voi4hkASctS3ie7KlNej0O1/t67kpeyDZTjUH/Ly/dZCT2tS6O1oUb6LovluV6N1Iby/Dn3l7zUudx2VLedvqamJh588MFaF6Ok5uZmJk+eXOti9Eo1yi7pP/v1F5Spv9tQHtpBrctQyd9/2mmn8aMf/Yj29namTJmyDTgI2BwRO1KWNmB0ej4a2AAQETskbUn5RwP3Fr1t8T4bOqQfV6ockuYB8wBGjhzJP/3TP/W5bpW2detW9t9//1oXo6S8lG3KlCkD4jjU32p9jKmkntal0f6XNdJ3WSzP9WqkNpTnz7m/5KXO5bajuu30mdnAsWLFCkaMGMGECRNobm4uJKtE1uhmW2fppULdS97PJiKuAq4CmDhxYuThgN9RXv4RlZLnspmZmTUqd/rMLPfWrFnD8uXLWbVqFdu3b4fsJs/fAYZJGpyu9o0hu1kvZFfqDgHaJA0GDgBeLUovKN6ns3QzMzOzuuaFXMws9xYuXEhbWxutra0sW7YMoD0iZgJ3A6elbLOBW9Pz5ek1aftdEREpfUZa3fNQYCxwP/AAMDatBroP2WIvy6tRNzMzM7P+5it9tpvHn9vCnAUru83XuugTVSiN9VZT+g7nj9/R7fdZ59/lV4Blkr4BPAxck9KvAa5PC7W8StaJIyKekHQz2QItO4BzImIngKRzgdXAIODaiHiiqjUZAJoWrBwIbbLhNRV9f919n/4urVqaOmmHHduo26T1VbnnWHlra+70mVldSfPBWgDSapvHdswTEduB00vtHxEXAxeXSF8FrKpgUc3MzMxyweGdZmZmZmZmDcydPjMzMzMzswbmTp+ZmZlZmXbu3MnRRx8N8EGAtADUfZLWS7opLQZFWjDqJkktaXtT4T0kXZDSn5J0clH61JTWImlBdWtmZo3MnT4zMzOzMi1evJhx48YVJ30LuCwixgKbgLkpfS6wKSI+CFyW8iHpMLLFpQ4HpgLflzRI0iDgCuAU4DDgjJTXzKzPuu30SXqXpPslPSrpCUn/mNI9smVmZmYDRltbGytXruTss88GQJKAjwK3pCxLgE+n59PSa9L2E1P+acCyiHgjIp4lW5jq2PRoiYhnIuJNYFnKa2bWZ+Ws3vkG8NGI2Cppb+AXkm4Dzicb2Vom6QdkI1pXUjSyJWkG2cjWZzuMbL0P+LmkD6XfcQXwcbIbJz8gaXlEPFnBepqZmZn1yXnnnccll1xCe3t7IekgYHNE7Eiv24DR6floYANAROyQtCXlHw3cW/S2xfts6JB+XKlySJoHzAMYOXIkzc3NfahVZuvWrRV5n1qZP35HyfSR++2+rZ7raNYX3Xb60g2Nt6aXe6dHkI1sfS6lLwEuIuv0TUvPIRvZ+l7HkS3g2XT/rMJS6y1p6XUkFUa23OlrEGeddRYrVqxgxIgR/PKXvwRA0kXAXwIvp2x/n5bMR9IFZIMHO4EvRsTqlD4VWEx2H7UfRsSilH4o2YjogcBa4Mw0StpwOrsPkZmZ9a/C/7EJEyYUdxxUImt0s62z9FLRV1EijYi4CrgKYOLEiZFuZdMnzc3NVOJ9aqWz+6XNH7+DSx9/53S3debkKpXILF/Kuk9fijN/iGzS8hXA0zTIyFZ/qOfRso4jYp3pSf2OPPJIJk2axMKFCzvud1lE/FNxQi+vCBfmU3S86mxmZlYRa9asYfny5axatYrt27cDDAG+AwyTNDidE40Bnk+7tAGHAG2SBgMHAK8WpRcU79NZuplZn5TV6YuIncBRkoYBPwXGlcqWftbVyFZ/qOfRsu/ecOtuI2Kd6clI2eTJk2ltbeXyyy8v53Pp0RVhSevo/KqzmZlZRSxcuJCFCxcC2f/5KVOmtEfETEn/ApxGFnEyG7g17bI8vb4nbb8rIkLScuDHkr5NNrg5Frif7DxpbIpeeY5sAPRzmJlVQFmdvoKI2CypGZiER7as786VNAt4EJgfEZvo+RXhruZT7KGaV4v744pvOVdhi5Vz5bbePgOz3tiwYQOzZs3ihRdeYK+99gIYASDpQOAmoAloBT4TEZvStITFwKnA68CciFib9pkN/EN6629ExJKUPgG4DtgPWAV8KU2RsMb2FWCZpG8ADwPXpPRrgOvT4OWrZJ04IuIJSTeTTWPZAZyTBteRdC6wmmwaw7UR8URVa2JmDavbTp+k3wPeSh2+/YCPkYXT3Y1Htqz3rgS+TnZV9+vApcBZ9PyKcFfzKfbcUMWrxf1xxbezOQud6TiXoZT+nN9Qz1e9rbEMHjyYSy+9lGOOOYb29naGDh06IoWTzwHujIhFafXoBWQn8aeQ/Z8aSzbAdCVwXOokXghMJDvWPJRCzTelPPPIBq5WkS3Hf1tVK2pVkY5rLQApAuXYjnkiYjtweqn9I+Ji4OIS6avI2o6ZWUWVc6VvFLAkzevbC7g5IlZIehKPbFkvRcSLheeSrgZWpJc9vSL8Cp1fdTYzA2DUqFGMGjUKgCFDhgBsI4sKmAZMTtmWAM1knb5pwNJ0pe5eScMkjUp574iIVwEk3QFMTVEwQyPinpS+lGzpfnf6zMys5spZvfMx4OgS6R7Zsl6TNCoiNqaXfw78Mj3v0RXhdBW5s6vOZmZ7aG1tBXg3cB8wsnAsioiNkkakbG8vSpYUQse7Sm8rkb6HelmUrDPFIePdhZDXU90cjm5mjaxHc/rMeuOMM86gubmZV155hTFjxgAcDFwi6Siy8KhW4K+g11eEO5tPYWa2m61btzJ9+nSADRHxWjZ1r6SeLkpWdqh5vSxK1pniMPPuQsjraXl8h6ObWSNzp8/63Y033rjba0mvRMSZneXv6RXhzq46m5kVe+utt5g+fTozZ85k7dq1m1Pyi4XIgxS++VJK7yzUvI13wkEL6c0pfUyJ/GZmZjVXanEMMzOzhhIRzJ07l3HjxnH++ecXbyosPgZ7Lko2S5lJwJYUBroaOEnScEnDgZOA1Wlbu6RJaeXPWTjU3MxK2LlzJ0cffTRk979G0qGS7pO0XtJNkvZJ6fum1y1pe1PhPSRdkNKfknRyUfrUlNaSFqcyA9zpMzOzAWDNmjVcf/313HXXXRx11FEAh0k6FVgEfFzSeuDj6TVkUQXPkK3QeDXwNwBpAZevAw+kx9cKi7oAfw38MO3zNF7ExcxKWLx4MePG7XbL628Bl0XEWGATMDelzwU2RcQHgctSPtLKwzOAw8lWCf6+pEFp0cUryFYfPgw4I+U1c3inmZk1vuOPP57iW+ZJejKFjAOc2DF/WrXznFLvFRHXAteWSH8QOKIiBTazhtTW1sbKlSv56le/yo033kiKDPgo79yubAlwEdktYKal5wC3AN9L+acByyLiDeDZtGJ+YZpLS5r2gqRlKe+T/V8zyzt3+sws97Zv384JJ5zAG2+8wY4dOyBb2ZW0musy4EBgLXBmRLwpaV9gKTAB+C3w2YhoTftcQDZ6uhP4YkSsTulTyW7GPQj4YUQswszMrILOO+88LrnkEtrb2wtJBwGb022nYPeVf99eLTgidkjakvKPJrsfKCX26bi68HGlylHpVYQH0uq3hRWL6231Ynf6zCz39t13X+666y72339/3nrrLfbZZ5+haZ7V+WQhMcsk/YCsM3clRSExkmaQhcR8tkNIzPuAn0v6UPo1V5CF97UBD6Qbbnt01MzMKmLFihWMGDGCCRMmFHcIulr5t6erBZeatlWVVYQH0uq3hRWM6231Ys/pM7Pck8T+++8PZCswkv3DC7KQmFtStiVkN8OGLJxlSXp+C3Bix5CYiHiWbO7VsenREhHPRMSbZFcPp/V3vczMbOBYs2YNy5cvp6mpiRkzZgAMAb4DDJNU6D0Ur/z79irCafsBwKt0vbpwqXQzX+kzs/qwc+dOJkyYQEtLC8BrZAtl1H1ITH/Ia5jN/PE7ug2HgfyFxJiZVcLChQtZuHAhkB3npkyZ0h4RMyX9C3Aa2YBjx1WEZwP3pO13RURIWg78WNK3yaJWxgL3kw2Ijk1TH54ji2z5HGa402dmdWLQoEE88sgjbN68meHDh78HGFciW92FxPSHvIbZzFmwsttwGMhfSIyZWT/7CrBM0jeAh4FrUvo1wPVpoZZXyTpxRMQTkm4mW6BlB3BOROwEkHQu2a1lBgHXRsQTVa2J5ZY7fWZWV4YNGwbQDkwihcSkq32lQmLaygyJoYt0MzOzikoDcy0AabXNYzvmiYjtwOml9o+Ii4GLS6SvIrvljNluPKfPzHLv5ZdfZvPmzQBs27YNYCiwDribLOQFSofEQFFITEqfkW54eyjvhMQ8QAqJSTfFnZHympmZmdU9X+kzs9zbuHEjs2fPZufOnezatQvgtYhYIelJHBJjZmZm1iV3+sws9z784Q/z8MMPv/1a0kZwSIyZVY/vF2pm9cydPjMzM6Mp3XvKSvP9Qs2snnlOn5mZmVk3fL9QM6tnvtJnZmZmVoZGvl9oXu/vWa7O7v/Z8d6g9VxHs75wp8/MzMysDI18v9C83t+zXHM6CU/ueG9Q3wfUBiqHd5qZWcM766yzGDFiBEccccTbaZIukvScpEfS49SibRdIapH0lKSTi9KnprQWSQuK0g+VdJ+k9ZJuSrf+sAZV6n6haVOp+4VS5v1Cu7qPqJlZn7jTZ2ZmDW/OnDncfvvtpTZdFhFHpccqgA4LbUwFvi9pkKRBZAttnAIcBpyR8kK2SMdlETEW2ES2iIc1EN8v1MzqmcM7zcys4Z1wwgm0traWm/3thTaAZ9P9Hgu3BmlJtwpB0jJgmqR1ZIt5fC7lWQJcRLaCozUI3y/UzOqZO30DRLlLcc8f388FMTPLl3MlzQIeBOZHxCZ6vtDGQXS+mMce+mMRjkrobCGMrnRcJKOjvNStHOUsZHLZZZe9/XzKlCm+X6iZ1Y1uO32SDiG7ueh7gV3AVRGxWNKBwE1AE9AKfCYiNqXliBcDpwKvA3MiYm16r9nAP6S3/kZELEnpE4DrgP3IDnZfSiEQZmZm/eVK4Otki2V8HbgUOIueL7TRWf6S+mMRjkrobCGMrnRcJKOjelo0o94XMjEz60o5c/p2kI1+jiObsHxOmsOwALgzzV+4M72GbK7D2PSYRwpvSZ3EC8lGRY8FLpQ0PO1zZcpb2G9q36tmZmbWuYh4MSJ2RsQu4GreuVrT04U2XqHzxTzMzMxqrttOX0RsLFypi4h2sknLo9n9pqMdb0a6NDL3kv0jHAWcDP+PvXuPs6uu73//epMIRZIYIiSOhBqsyAkYwSQ/gkdKJ0Uhgm3wkaLElCSSnrRqLDzI6Y9Q+hN/3gj2pBaslx8VSoKaQLXWlIsxAvNr5RjkYiRCoIw4x0whXIRAwkWT+Dl/rO8OO5M9M3tm39Ze8yXFLaoAACAASURBVH4+Hvsxe3/X2nt/19qfWWt91/fGxoh4NjWf2QjMScvGRcSPUu3emrLPsgLoZ9S8CZI2ppHuNpZuAChzdRoZ7wFJ08vesyit/2iqNS6lz5C0Jb3n6lTbbGY2oHT+KXk/8LP0fEgDbaRzV3+DeZiZmbXckPr0SZoCvAO4G5gUEaX27E9ImphW2zcZaVLq2zBQem+F9Erfn8t+EH3lcYLTavtqDNY/o2Qo23fiiSdyyimncMUVV5S/r1RTvDINe74CuIT9a4pnkdUCzyqrKZ5J1mzqPknr0w2EUk3xJrLmwXOA26rO4AhXbX/PnpVnNzgnZo0zf/58urq6eOaZZ5g8eTLAEcDnJZ1EdkzpAf4chj3QxiVUHszDzMys5aou9EkaA3wbuCgiXhigMmWok5FW3Rcir/0g+spjv4Bq+2oM1j+jZCj9NDo7O+np6eHqq68u3y9zgdKL1UAX2UXTvppiYJOkUk1xJ6mmGEBSqaa4i1RTnNJLNcUu9JnZPmvXrt3vtaRnIuL8/tYf6kAb/Q3mYWZmlgdVFfokvYaswPeNiPiXlPykpI5Uy9cBPJXSB+oL0dknvSulT66wvhVb02uKobm1xY2o8R3q6HrV1txWYzjbksdabzMzM7ORpprRO0XWTGVrRPxd2aLSpKMrOXAy0mVp/qJZwPPpon4D8LmywVvOAC6NiGcl7ZR0Clmz0YXAF+uwbdaeGlZTDM2tLW5Eje9QR9ertua2GsMZhS+Ptd5mZmZmI001V4PvAs4HtkjanNL+mqywd5OkJcAveXUumlvJpmvoJpuy4cMAqXD3abKO8ACfKjXVAz7Cq1M23Iab5o0Erik2MzMzM2uCQQt9EfFDKtemAJxeYf0APtbPZ10HXFch/V7gbQe+wwrMNcVmZmZmZk1Qn3ZfZgPoZ9Q81xSbmZmZmTWBC33WcP2MmvcrXFNsVdq2bRsLFy5k+/btHHTQQQATIZvvEbgRmEI25P4HIuK51Bf5KrIbCC8Bi0vzjaY5Hv8mffRnImJ1Sp/BqzcPbgUuTPFoZmZm1tYGnZzdzKzVRo8ezapVq9i6dSubNm0CmCjpeF6d7/FY4Pb0Gvaf73Ep2VyOpULi5WRNh08GLi9rMlya77H0vjnN2DYzMzOzRnOhz8xyr6Ojg+nTpwMwduxYgJfJpuaYSzbPI+nvOen5vvkeI2ITUJrv8UzSfI8R8RxQmu+xgzTfY6rdW1P2WWZmZmZtzc07zayt9PT0ALyWbOCeps/32My5Hocrr/MjLp+2p6q5I/OYdzMzs3bmQp+ZtY1du3Yxb948gG0R8ULWda+ihs332My5Hocrr/MjLl5xS1VzRw5nTkgzMzPrn5t3mllb2L17N/PmzWPBggUAO1Lyk6lpJkOY77G/dM/3aGb92rZtG7Nnz2bq1KmccMIJUDaglKSNkh5Nfw9P6ZJ0taRuSQ9Iml76LEmL0vqPpsGlSukzJG1J77laA9zZMjMbChf6zCz3IoIlS5YwdepULr744vJFpfke4cD5Hhemi65TSPM9AhuAMyQdni7MzgA2pGU7JZ2SLrIWln2WmZkHlDKztubmnWaWe3fddRc33HAD06ZN46STTgI4XtJZeL5Hs1ybsuKWqtbrWXl2g3NSu46ODjo6OoCKA0p1ptVWA13AJZQNKAVsklQaUKqTNKAUgKTSgFJdpAGlUnppQCkfi8ysZi70mVnunXrqqZRPmSfpoYi4Nb30fI9WlQsuuICbb76ZiRMn7kvzXI82HEUcUCqvA0BVq78BovoOHtXO22hWCxf6zHKg2rvhZjZ8ixcvZtmyZSxcuLA8udQ0b6WkFen1JezfNG8WWbO7WWVN82aSDfZzn6T1aQqQUtO8TWSFvjm4lqZwijqgVF4HgKrW4n7Oo30Hj/JAUTZSudBnZmY1aZebFqeddlqphqacm+ZZ1coHlLr//vv3G1Aq1fJVO6BUZ5/0LjyglJk1kAdyMTOzkWy/pnmkERlpYNM8a08eUMrM2plr+szMzA7UsKZ5jeiPVQ/99YkaSN/+UsOVh30wWJ+2LVu2cMMNN/DmN7+Zm2++GTyglA3Rtm3bWLhwIdu3b+eggw6Csmk/cN9iazAX+szMbCRretO8RvTHqof++kQNpG9/qeHKQz+rwfq0dXZ28vGPf3zfaw8oZUNVmvZj+vTp7Ny5k3HjxpWm/ViM+xZbg7l5p5mZjWRummdmTdHR0cH06dOBitN+rE6rrSbrDwxlfYsjYhNQ6lt8JqlvcSrolfoWd5D6FqebDmvKPstGONf0mZnZiDB//ny6urp45plnAN6emuO5aZ6ZNV3Rpv1o9yk/hqLUpH2w5u152x8u9JmZ2Yiwdu3afc8lPRAR16aXbppnZk1TxGk/2n3Kj6EoNYUfrHl7Hpqtl3PzTjMzMzOzJiif9gPYb9oPgCH0Le4v3dN+WEUu9JmZmZmZNZin/bBWcvNOMzMzM7MGu+uuu7jhhhuYNm0aJ510EnjaD2siF/rMzMzMzBrs1FNPpXzKPE/7Yc3k5p1mZmZmZmYFNmihT9J1kp6S9LOytAmSNkp6NP09PKVL0tWSuiU9IGl62XsWpfUflbSoLH2GpC3pPVdrgCGMzMzMzMzMbGiqqem7HpjTJ20FcHtEHAvcnl4DvBc4Nj2WAl+BrJAIXA7MAk4GLi8VFNM6S8ve1/e7rMAk9aRC/2ZJ96a0ut1UMDMzMzMb6QYt9EXEvwPP9kmeC6xOz1cD55Slr4nMJmB8Gnr2TGBjRDwbEc8BG4E5adm4iPhRare8puyzbOSYHREnRcTM9LqeNxXMzMzMzEa04Q7kMikNC0tEPCFpYko/CthWtl5vShsovbdCekWSlpJd7DNp0qTczXRfsmvXrtzlbfm0PVWtN+nQ6tZt8PbNBTrT89VAF3AJZTcVgE2SSjcVOkk3FQAkbSSrMV6LmZmZmdkIV+/ROyv1x4thpFcUEdcA1wDMnDkzOjs7h5HFxuvq6iJveVu84paq1ls+bQ+rtgweFj0LOmvM0T4BfF9SAP8r/cb1uqlwgGbeOBhK4b/aQvlQVVuIr8Zw9lW9boBceeWVbNq0ifHjx/NP//RPwL4a3huBKUAP8IGIeC71C76KbJjrl4DFEXF/es8i4G/Sx34mIlan9Bm8OsT1rcCFUT7EmpmZmVkbG26h70lJHemCvAN4KqX3AkeXrTcZeDyld/ZJ70rpkyusbyPHuyLi8VSw2yjp4QHWrfnmQTNvHAyl8F9toXyoqi3EV2M4Bf163QA56KCDGDNmDAsXLiz/vFIz4JWSVqTXl7B/M+BZZM2AZ5U1A55JFh/3SVqfmpyX+hZvIiv0zcFzG5mZmVlBDHfKhvVAabCMRcB3y9IXpgE3TgGeTzU2G4AzJB2e+lqdAWxIy3ZKOiXdnV9Y9lk2AkTE4+nvU8B3yPrkPZluJjCEmwqV0q0gTjvtNCZMmNA32X2LzczMzKpQzZQNa4EfAcdJ6pW0BFgJvEfSo8B70mvI7pA/BnQD/wh8FCD1tfo0cE96fKrU/wr4CPC19J6f47vrI4akwySNLT0nuxnwM+p0U6GJm2KtsV8zYKApfYvNbOS64IILmDhxIm9726tzX3saKzNrB4O2+4qI+f0sOr3CugF8rJ/PuQ64rkL6vcDbDnyHjQCTgO+kc9po4JsR8T1J9wA3pRsMvwTOTevfStZPq5usr9aHIbupIKl0UwH2v6lgI09D+xa3w4BSzR5Maij9RqvpZ5rHfWoGsHjxYpYtW8bChQvLk93U3Mxyr94DuZhVLSIeA06skP4r6nRTwQqtJX2L22FAqWYPJjWUPqnV9DOt40BRVZPUA+wE9gJ7ImJmPQcLsmI47bTT6Onp6ZtclxGnJXWRmpqn9FJTcxf6zKxmLvSZWbsqNQNeyYHNgJdJWkd2d/35VDDcAHyubA7HM4BLU03xztRk+G6yvsVfbOaGWG7Mjohnyl7XswbHiqvp01g1osVBHqebGor+WhD0bV3QzttoVgsX+sws9+bPn09XVxfPPPMMkydPBjiCrLBXr2bAH+HVKRtuw3fWLeM5Q60WDWtq3ogWB3mcbmoo+mtx0Ld1QStaEpjlgQt9ZpZ7a9fuf80s6Zl6NgN232KjiXOG5rVf6HDm9KzXXKB52AfV1nRt376dF198sTzJ01iZWe650GdmZtbEOUPz2i90OPOF1msu0DzUvlRb09XT08Nhhx1WnuSm5m1kSpVx3rPy7AbnxKy5XOgzM7MRr3zOUEn7zRlahxocK4jypubA28umsXJTczPLNRf6zMxsREvzhB4UETvL5gz9FHWqwWniplRUbc2GDa68qbmkByLi2vTSTc3NLNdc6DMzs5HOc4aamVmhudBnZlVxPwgrKs8ZamZmRXdQqzNgZmZmZmZmjeNCn5mZmZmZWYG50GdmZmZmZlZg7tNn1kAeNc/MzMws/4p+zeZCn5mZmZnlUtEvxM2axc07zczMzMzMCsyFPjMzMzMzswJzoc/MzMzMzKzAXOgzMzMzMzMrMA/k0ubcwdnMzMzMzAbimj4zMzMzM7MCc6HPzMzMzMyswNy808zqqrzJ8fJpe1jcTxPknpVnNytLZmZmZiNabgp9kuYAVwGjgK9FxMoWZ8najGPIauUYslo1M4aK1Ke72m0ZCTeLfByyWjmGrJJcNO+UNAr4EvBe4HhgvqTjW5srayeOIauVY8hq5RiyWjmGrFaOIetPXmr6Tga6I+IxAEnrgLnAQy3NlbWTpsZQNXell0/bQ37+xawKPg710aqapDau9XEMWa0cQ1Yrx5BVlJcr0qOAbWWve4FZfVeStBRYml7ukvRIE/I2HEcAz7Q6E8Pxl1XmXVfW9DVvqundleUuhqrdl43U6jwM9P2OoYZqeez1p54x6RhqvWYfY2r8zQcz1G0pWgzl9rhRi+HGaINjraRIMVTI+BnIYLHVpBiCKuMoL4U+VUiLAxIirgGuaXx2aiPp3oiY2ep8DEcb5z13MZSHfdnqPLT6+4codzE0XHne73nOWx0UJoaqVaTfMyfb0rIYysn2111Rt2sALYmhEbif226bc9Gnj+wuxNFlrycDj7coL9aeHENWK8eQ1coxZLVyDFmtHENWUV4KffcAx0o6RtLBwHnA+hbnydqLY8hq5RiyWjmGrFaOIauVY8gqykXzzojYI2kZsIFseNnrIuLBFmerFu3cbKct857TGMrDvmx1Hlr9/VXLaQwNV573e57zVpOCxVC1ivR7tnxbWhxDLd/+BinqdlXUwhgaUfs5aattVsQBzXzNzMzMzMysIPLSvNPMzMzMzMwawIU+MzMzMzOzAnOhr04knSvpQUm/lTSzz7JLJXVLekTSma3K40AkzUn565a0otX5aUeS/lbSw5IekPQdSeP7Wa9H0hZJmyXdW6fvHvD3k3SIpBvT8rslTanH96bPPlrSnZK2pv+BCyus0ynp+bTNmyV9ol7fb6/K83HIx5hiKdrv2YjjcrvI83GjFkWL0bwa6NqnneNnMG0ZXxHhRx0ewFTgOKALmFmWfjzwU+AQ4Bjg58CoVue3T95HpXy9GTg45ff4Vuer3R7AGcDo9PxK4Mp+1usBjmjm7wd8FPhqen4ecGMdv78DmJ6ejwX+s8L3dwI3t/o3Kvojr8chH2OK9Sji71nv43I7PfJ63KhxmwoXo3l99Hft087xU8U2t2V8uaavTiJia0Q8UmHRXGBdRPw6In4BdAMnNzd3gzoZ6I6IxyLiN8A6snzbEETE9yNiT3q5iWxunGao5vebC6xOz78FnC6p0gSuQxYRT0TE/en5TmArcFQ9PtuGJsfHIR9jisW/Z4Hk+LhRC8dokwxw7dPO8TOYtowvF/oa7yhgW9nrXvJ3QdwOeWw3FwC39bMsgO9Luk/S0jp8VzW/37510sH5eeD1dfju/aRmo+8A7q6w+J2SfirpNkkn1Pu7bUCt/h9v9fdbfRXx96z3cbkI2vl3bue8t7Pya58i/wZtuW25mKevXUj6AfCGCosui4jv9ve2Cml5myejHfKYC9XEgKTLgD3AN/r5mHdFxOOSJgIbJT0cEf9eS7YqpPX9/Rr+G0saA3wbuCgiXuiz+H7gTRGxS9JZwL8Cx9bz+0eKNj0Otfr7rb6K+HvW+7icK2163KhFO+c9d4Z57VPk36Att82FviGIiHcP4229wNFlrycDj9cnR3XTDnnMhcFiQNIi4H3A6ZEaflf4jMfT36ckfYesmUAtFxfV/H6ldXoljQZeBzxbw3fuR9JryAp834iIf+m7vLwQGBG3SvqypCMi4pl65WGkaNPjUKu/3+qrcL9nA47LudKmx41atHPec2eY1z5F/g3actvcvLPx1gPnpdETjyGr3fhxi/PU1z3AsZKOkXQw2UAf61ucp7YjaQ5wCfDHEfFSP+scJmls6TlZB+if1fjV1fx+64FF6fmfAHf0VygdqtQ38Fpga0T8XT/rvKHUh1DSyWTHnl/V4/utKq0+DvkYUyyF+j0bdFwuglYfN2pRqBjNswGufdo5fgbTlvHlmr46kfR+4IvAkcAtkjZHxJkR8aCkm4CHyKq9PxYRe1uZ174iYo+kZcAGshGJrouIB1ucrXb0D2SjVG1M5ZtNEfEXkt4IfC0izgImAd9Jy0cD34yI79Xypf39fpI+BdwbEevJCmU3SOomq+E7r5bv7ONdwPnAFkmbU9pfA7+b8vdVsoLmRyTtAV4GzqtXodNeldfjkI8xxVLA37Pux+V2ktfjRi0KGKN5VvHap53jZzDtGl/ydZeZmZmZmVlxuXmnmZmZmZlZgbnQZ2ZmZmZmVmAu9DWJpA2pj1Xf9LmStkt6j6Q7JT0vqafCeidJ+o+0vFfSJ5qSccuNOsTQ/ynpx5J2SnpA0qlNybjlVhUx9VeSfpZi5heS/qoV+bT8qDVmJE1Jx6mXJD0saTijSlobq0MMfVrSFkl7JH2yaRm33KglhiRNlLRW0uPpeukuSbOauwWt4UJf81wPnF8awbDM+WRzmjwPXAf0d1H1TbLhoycAf0A2KMYfNyarllPXM8wYkjSBbGSpvwXGA58H/k3S4Y3MsOXe9QwcUwIWAocDc4Blkuo5CJC1n+upLWbWAj8BXg9cBnxL0pGNzrTlyvXUFkPdwH8Hbml8Vi2nrmf4MTSGbPTNGWTX1KvJBjAa04R8t5QHcmkSSYcC24E/Kk34mi64nwBmRcRPU9q7yUZ6nNLn/S8BMyPiofT6n4H7I+KK5m2FtVItMSTpfcCVEXFCWdp/prRrm7cVlifVxlTZ+leTnTc+3vTMWi7UEjOS3gpsAY6IiJ1p+X+Qze/51WZuh7VOvY47kr4OdEfEJ5uSccuNep+7JL0AzI6I+xqb89ZyTV+TRMTLwE1kdx5KPgA83Dc4+/H3wEJJr5F0HPBO4Af1z6nlVY0xpPTom/a2+uXQ2s1QYirdUf19IPfDUlvj1BgzJwCPlQp8yU9Tuo0QPu5YreoZQ5JOAg4mq0EuNBf6mms1cG66QwFZsK6u8r03k8119jLwMHBtRNxT/yxazg03hv5f4I2S5qcbB4uA3wNe26B8WvuoNqY+SXbO+Kcm5cvya7gxM4asGXq554GxDcij5ZuPO1armmNI0jjgBuB/RkTfY1PhuNDXRBHxQ+BpYK6kNwP/jayv3oBSf6zvAZ8Cfgc4GjhT0kcbmF3LoeHGUET8CpgLXAw8SdbG/QdAb+Nya+2gmphSNgntQuDsiPh183NpeVJDzOwCxvX5uHHATmxE8XHHalVrDKXC4r+RTSY/IrpKjW51BkagNWQBeBzw/Yh4sor3vBnYGxFr0uteSeuAs4AvNyablmPDiSEi4n+THRSRNBr4ObCqUZm0ttJvTEm6AFgBnBYRvklgJcOJmQeBN0saW9bE80SquHFlheTjjtVqWDEk6RDgX4H/Av68edltLdf0Nd8a4N3A/0VZNbSkgyT9DvCa7KV+R9LBafF/prQPpfXeAHyQrC+EjTzDiSEkvSM17RwH/D9Ab0RsaHLeLZ/6i6kFwOeA90TEYy3Km+XTkGMmIv4T2Axcno5P7wfeDny7abm2PBnWcSedx36H7Bp2dIqlUU3Ks+XLkGNI0muAb5F1l1oYEb9tXnZby6N3toCkLrK7m28oVTdL6gTu7LPq/46IzrT8D4ErgbeSBeq/ARdGxEvNybXlyTBjaC1Z7TBkzYU/HhFPNSO/ln/9xNQvgMlAebOYr0fEXzQ/h5Y3w4kZSVPIhlufBfwS+FhEeFCyEWqYMXQ9sKjPR304Iq5vdH4tf4YaQ5L+AOgiu5YuL/C9NyL+oymZbhEX+szMzMzMzArMzTvNzMzMzMwKzIU+MzMzMzOzAnOhz8zMzMzMrMBc6DMzMzMzMyuwtp2nb/z48fGWt7yl1dnYz4svvshhhx3W6mwcIG/5uu+++56JiCNbnY8jjjgipkyZUvfPbdX+Hknfm6cYOvLII3P1/5VHeTsGQb5iqNJxKI/7rJ6KsH15j6FK8r7f854/qG8e2zGG6qEVv3ORr5GqjqOIaMvHW9/61sibO++8s9VZqChv+QLujRzE0IwZMxqyfa3a3yPpe/MUQ3n7/8qjPO6jPMVQJXncZ/VUhO3LewxVkvf9nvf8RdQ3j+0YQ/XQit+5yNdI1caRm3eamZmZmZkVmAt9ZmZmZmZmBeZCn5mZmZmZWYG17UAulpmy4pZB11k+bQ+djc+KNcFAv/fyaXtYnJb3rDy7WVkyG/Q4VIpNx+XI019slB+vwMeskahSbPSNC3BsWO2mrLilYmz1VfRYc02fmZmZmZlZgbnQZ2ZmZmZmVmAu9JmZmZmZmRWYC31mZmZmZmYF5kKfmZmZmZlZgbnQZ2ZmZmZmVmAu9JmZmZmZmRWYC31mZmZmZmYF5kKfmZmZmZlZgbnQZ2ZmZmZmVmAu9JmZmZmZmRWYC31mZmZmZmYF5kKfmZmZmVmDvfLKK5x88smceOKJnHDCCQBvBJB0jKS7JT0q6UZJB6f0Q9Lr7rR8SumzJF2a0h+RdGZZ+pyU1i1pRXO30PJsdKszYM0xZcUtVa3Xs/LsBufEzMzMbOQ55JBDuOOOOxgzZgy7d+/m4IMPHifpFOBi4AsRsU7SV4ElwFfS3+ci4i2SzgOuBD4o6XjgPOAEsoLjDyS9NX3Nl4D3AL3APZLWR8RDzd1SyyPX9JmZmZmZNZgkxowZA8Du3bsBBATwh8C30mqrgXPS87npNWn56ZKU0tdFxK8j4hdAN3ByenRHxGMR8RtgXVrXzDV9ZmZmZmbNsHfvXmbMmEF3dzfAC8DPgR0RsSet0gsclZ4fBWwDiIg9kp4HXp/SN5V9bPl7tvVJn1UpH5KWAksBJk2aRFdXV20bNgS7du1q6vctn7aHSYdmfwfSiDw1e1sH4kKfmbWNvXv3MnPmTIC3QNYPguxO5gTgfuD8iPiNpEOANcAM4FfAByOiJ73nUrImM3uBv4yIDSl9DnAVMAr4WkSsbOKmmZnZCDBq1Cg2b97Mjh07OPzwww8DplZYLdJf9bOsv/RKLfiiQhoRcQ1wDcDMmTOjs7NzsKzXTVdXF838vsUrbmH5tD2s2jJwsadnQWfdv7vZ2zoQN+80s7Zx1VVXMXXqfufHK8n6QRwLPEdWmIOyfhDAF9J69OkHMQf4sqRRkkaR9YN4L3A8MD+ta2ZmVnfjx48H2AmcAoyXVCqRTAYeT897gaMB0vLXAc+Wp/d5T3/pZi70mVl76O3t5ZZbbuHP/uzPAEj9GtwPwszM2sLTTz/Njh07AHj55ZcBxgFbgTuBP0mrLQK+m56vT69Jy++IiEjp56XRPY8BjgV+DNwDHJtGAz2Y7Cbn+kZvl7UHN+80s7Zw0UUX8fnPf56dO3eWkl5PDvpB5Km9fqsM1k+i1JdipO8nMxvZnnjiCRYtWsTevXv57W9/C/BCRNws6SFgnaTPAD8Brk1vuRa4QVI3WQ3feQAR8aCkm4CHgD3AxyJiL4CkZcAGsq4K10XEg03cRMsxF/rMLPduvvlmJk6cyIwZM8oLDv31aRhoWd37QYwZMyY37fVbZfEgU8KU+lI0or+EmVm7ePvb385PfvKTfa8lPQEQEY+RtTjZT0S8Apxb6bMi4rPAZyuk3wrcWqcsW4G4eac13LZt25g9ezZTp04tTUY6EUDSBEkb02SkGyUdntIl6eo0segDkqaXPkvSorT+o5IWlaXPkLQlvefq1JTPCuKuu+5i/fr1TJkyhfPOOw9gLPD3uB+EmZmZ2aBc6LOGGz16NKtWrWLr1q1s2rQJYGIaJGMFcHsahOP29BqywTSOTY+lZBOUImkCcDlZs7uTgctLBcW0ztKy981pxrZZc1xxxRX09vbS09PDunXrAHZGxALcD8LMzMxsUC70WcN1dHQwfXpWWTd27FiAl8n6UZUPttF3EI41kdlEVpvTAZwJbIyIZyPiOWAjMCctGxcRP0oX9mvKPsuK7RLg4tTf4fXs3w/i9Sn9YtINhdS3odQP4nukfhCpX2CpH8RW4Cb3gzAzM7OiGLTQJ+loSXdK2irpQUkXpnQ3zbMh6+npAXgtcDcwKSJK7dmfIDX7pGwQjqQ02MZA6b0V0q2AUv+5bsj6QUTEyRHxlog4NyJ+ndJfSa/fkpY/Vnp/RHw2In4vIo6LiNvK0m+NiLemZQf0k7Bi2Lt3L+94xzugbK5HSXen89KNqaaXVBt8Yzov3S1pSukzJF2a0h+RdGZZ+pyU1i1pBWZmZjlRzUAue4DlEXG/pLHAfZI2AovJmuatTCe3FWR33cub5s0ia3Y3q6xp3kyyARLuk7Q+1diUmuZtIut8Oge4DSuUXbt2MW/ePIBtEfHCAGX7oQ7CMdCAHvt/cJ+RFxsxmmAjR3McaJTE0giJQFNHSfToldZOSnM9bt68uZRUmutxnaSvks3x+BXK5nqUdF5a74N95np8I/ADSW9Nn/Ul4D1kN57uj6+GDgAAIABJREFUSee4h5q2cWZmZv0YtNCXamBKtTE7JW3l1aZ5nWm11UAXWaFvX9M8YJOkUtO8TlLTPIBUcJwjqYvUNC+ll5rmudBXILt372bevHksWLCA+++/f0dKflJSR0Q8kWLkqZQ+0GAbnX3Su1L65ArrH6DvyIuNGHWxq6urYaM5DjRKYmmERKCpoyQ2cnvN6qk01+Nll13G2rVry+d6/FBaZTXwSbJC39z0HLK5Hv+h71yPwC9SE+LSqHvdpVplSaW5Hl3oK5i9e/cyc+ZMKKstJpvbcwJwP3B+RPxG0iFk3Q1mAL8CPhgRPek9l5LdWNgL/GVEbEjpc4CryIbb/1pErGzipplZgQ1pyobUvOUdVGiaJ6nhTfPKa2mOPPLI3NUutKLGY7D5sWD/GqDBNCL/EcEVV1zBuHHj9vXtS0qDbazkwEE4lqWLplnA8ynGNgCfKxu85Qzg0oh4VtJOSaeQxeZC4It13xAza2t5neux0nG3KDXo/Z17+p6X2mlbb7rpJiZMmFCe5NpiM8u9qgt9ksYA3wYualXTvPJamuOOO64htTS1aEWNx2DzY8H+NUCDaUQN0Q9/+EM2btzItGnTuOiiiwCOl3QWWWHvJklLgF/y6lw0twJnkfXbegn4MEAq3H2abKRFgE+Vao6BjwDXA4eS1RK7ptjM9snzXI+VzhtFqUHv7xzV97zULnM49vb28sgjj3DZZZdxxx13uLbYzNpGVSUBSa8hK/B9IyL+JSU3vWmetadTTz2VrLVvRtJDafJQgNP7rp+aBn+s0mdFxHXAdRXS7wXeVpcMm1nhlOZ6vPXWW3nllVegz1yPqbav0lyPvVXO9cgA6VYQ7VRbXEmeapAr1QJXapmUl/yW5Gkfmg3FoIW+dFfqWmBrRPxd2SI3zTMzs7ZwxRVXcMUVVwDZReTs2bN3RsQCSf9MNpfjOirP9fgjyuZ6lLQe+KakvyNrmlea61GkuR6B/yJrvvchrDDarba4kjzVIFeqBa7UMilvtcB52odmQ1FNTd+7gPOBLZJKw539NW6aZ2Zm7e8SYJ2kzwA/Yf+5Hm9ITe+eJSvEEREPSirN9biHNNcjgKTSXI+jgOs812OxuLbYzNpZNaN3/pDKd6XATfPMzKzN9J3rkVf7U+0TEa/w6s3Mvss+Cxwwl2Nqtn7rge+wInBtsZm1syGN3mlmZmZm+3FtsZnlngt9ZmZmZkPg2mIzazeVOg2bmZmZmZlZQbjQZ2ZmZmZmVmAu9JmZmZmZmRWYC31mZmZmZmYF5kKfmZmZmZlZgbnQZ2ZmZmZmVmAu9JmZmZmZNdi2bduYPXs2U6dO5YQTTgCYCCBpgqSNkh5Nfw9P6ZJ0taRuSQ9Iml76LEmL0vqPSlpUlj5D0pb0nqslqdnbafnkQp+ZmZmZWYONHj2aVatWsXXrVjZt2gQwUdLxwArg9og4Frg9vQZ4L3BseiwFvgJZIRG4HJhFNkfk5aWCYlpnadn75jRj2yz/PDm7mZmZmVmDdXR00NHRAcDYsWMBXgaOAuYCnWm11UAXcElKXxMRAWySNF5SR1p3Y0Q8CyBpIzBHUhcwLiJ+lNLXAOcAtzV+65pvyopbWp2FtuJCn5mZmZlZE/X09AC8FrgbmBQRTwBExBOSJqbVjgK2lb2tN6UNlN5bIf0AkpaS1QgyadIkurq6atqeodi1a1ddvm/5tD1Vrzvp0MHXb8Q+qNe21oMLfWZmZmZmTbJr1y7mzZsHsC0iXhig212lBTGM9AMTI64BrgGYOXNmdHZ2DpLr+unq6qIe37d4CDV9y6ftYdWWgYs9PQs6a8zRgeq1rfXgPn1mZmZmZk2we/du5s2bx4IFCwB2pOQnU7NN0t+nUnovcHTZ2ycDjw+SPrlCuplr+syKaCjt3HtWnt3AnJiZmRlARLBkyRKmTp3KxRdfzPLly0uL1gOLgJXp73fL0pdJWkc2aMvzqfnnBuBzZYO3nAFcGhHPStop6RSyZqMLgS82Z+ss71zoMzMzMzNrsLvuuosbbriBadOmcdJJJwEcL+ksssLeTZKWAL8Ezk1vuRU4C+gGXgI+DJAKd58G7knrfao0qAvwEeB64FCyAVwKOYiLDZ0LfWZmZmZmDXbqqaeSDcSZkfRQRNyaXp7ed/00aufHKn1WRFwHXFch/V7gbXXJsBWK+/SZmZmZmZkVmAt9ZmZmZmZmBeZCn5mZmZmZWYG50GdmZmZmZlZgLvSZmZmZmZkVmAt9ZmZmZmZmBeZCn5mZmZmZWYG50GdmZmZmZlZgLvSZWe5t27aN2bNnM3XqVE444QSAiQCSJkjaKOnR9PfwlC5JV0vqlvSApOmlz5K0KK3/qKRFZekzJG1J77lakpq9nWZmZmaN4EKfmeXe6NGjWbVqFVu3bmXTpk0AEyUdD6wAbo+IY4Hb02uA9wLHpsdS4CuQFRKBy4FZwMnA5aWCYlpnadn75jRj28zMzMwazYU+M8u9jo4Opk/PKuvGjh0L8DJwFDAXWJ1WWw2ck57PBdZEZhMwXlIHcCawMSKejYjngI3AnLRsXET8KCICWFP2WWZmZmZtbXSrM2DFd8EFF3DzzTczceJEfvaznwH7alxuBKYAPcAHIuK51KTuKuAs4CVgcUTcn96zCPib9LGfiYjVKX0GcD1wKHArcGG6cLcC6unpAXgtcDcwKSKeAIiIJyRNTKsdBWwre1tvShsovbdC+gEkLSWrEWTSpEns2rWLrq6umrap3S2ftmfA5ZMOzdYZ6fvJzMysVQYt9Em6Dngf8FREvC2l+YLdqrZ48WKWLVvGwoULy5NLzfJWSlqRXl/C/s3yZpE1uZtV1ixvJhDAfZLWp9qaUrO8TWQxNAe4rSkbZ021a9cu5s2bB7AtIl4YoNtdpQUxjPQDEyOuAa4BmDlzZowZM4bOzs5Bcl5si1fcMuDy5dP2sGrLaHoWdDYnQxVs27aNhQsXsn37dg466CAo6xeKz2dmZlZw1TTvvJ4D+7a4H41V7bTTTmPChAl9k90sz4Zk9+7dzJs3jwULFgDsSMlPphgg/X0qpfcCR5e9fTLw+CDpkyukW0G4X6jVygNKmVk7G7SmLyL+XdKUPslzgc70fDXQRVZLs++CHdgkqXTB3km6YAeQVLpg7yJdsKf00gW7a2mKr+nN8uDApnmNaG7WyOZ+AzWjKzWhG6pa89qM5o0RwRVXXMG4ceP29e1L1gOLgJXp73fL0pdJWkd2cf58irMNwOfKLtLPAC6NiGcl7ZR0Clmz0YXAFxu6UdZUHR0ddHR0ABX7hXam1Xw+s36VbhxMnz6dnTt3Mm7cuNKNg8W45YqZ5dxw+/S1/IL9yCOPzF3/kFb07anmIn8ohYFG5X/79u28+OKL1Xx+w5rlwYFN8xrRLK+rq2vIzf2mDNI87lX9/8uWmtANVa1N7oazvUP1wx/+kI0bNzJt2jQuuugigOMlnUVW2LtJ0hLgl8C56S23kjXL6yZrmvdhgFS4+zRwT1rvU6WLd+AjvNo07zZ8oVVYeesXWum4WJS+ov2de/qel9ppW8vy6hsHZtY26j2QS9Mu2I877riGXLDXohkXv30N1pcGhlYYaFSfm56eHg477LDy/fOkpI50kVVts7zOPulduFneiHDqqadS3jVK0kMRcWt6eXrf9dNF1scqfVZEXAdcVyH9XuBtdcmw5VYe+4VWOm+04nzSCP2do/qel1rZ33M42uHGQSV5uplQ6YZApZvUeclvSZ72odlQDLfQ5wt2q5Wb5ZlZU5X3C73//vv36xfq85lVq11uHFSSp5sJlW4IVLpJnbcbAnnah2ZDMdx5+koX7HDgBfvC1Hn5FNIFO7ABOEPS4emi/QxgQ1q2U9IpqbPywrLPsoKYP38+73znO3nkkUeYPHkywBFkhb33SHoUeE96DVmzvMfImuX9I/BRyJrlAaVmefdwYLO8r6X3/Bw3hTGzPiKCJUuWMHXqVC6++OLyRT6fWdU8oJSZtatqpmxYS3ZX8whJvWSdj92Pxqq2du3a/V5LeiYifoWb5ZlZk9x1113ccMMNTJs2jZNOOgncL9SGqO+Ng+XLl5cWueWKmeVeNaN3zu9nkS/YzcysLbhfqNXKNw6sHi644AJuvvlmJk6cuC/N84VaM9R7IBczMzOzwvGNA6uHxYsXs2zZMhYuXFieXJov1NN+WMMMt0+fmZmZmZkNwWmnncaECRP6Js8lm+6D9PecsvQ1kdkElKb9OJM07Ucq6JWm/eggTfuRbjqsKfssG+Fc6DMzMzMza539pv0AmjL/tY0sbt5pZmZmZpY/DZv2Y7hzPdZDveY6rDTXY38qzQHZVyP2QZ7mdXShz8zMzMysdZo+X+hw53qsh3rNdVhprsf+VJoDsq9GzAmZp3kd3bzTzMzMzKx1PF+oNZxr+szMzMzMmmD+/Pl0dXXxzDPPALw9TfXhaT+s4VzoMzMzMzNrgrVr1+57LumBiLg2vfS0H9ZQbt5pZmZmZmZWYC70mZmZmZmZFZgLfWZmZmZmZgXmQp+ZmZmZmVmBudBnZmZmZmZWYC70mZmZmZmZFZgLfWZmZmZmZgXmQp+ZmZmZmVmBudBnZmZmZmZWYKNbnQEzMxsZpqy4par1elae3eCcmJmZjSyu6TMzMzMzMyswF/rMzMzMzMwKzM07c6jaJlBmZmZmZmaDcaHP9uM+NyOPf3MzMzOzYnOhz6yBXGtrZmZmZq3mPn1mZmZmZmYF5kKfmZmZmZlZgbnQZ2ZmZmZmVmAu9JmZmZmZmRWYC31mZmZmZmYF5kKfmZmZmZlZgeVmygZJc4CrgFHA1yJiZYuzZG3GMWS1cgxZrRxDw+P5Ql/lGLJaOYaGp+jHoVwU+iSNAr4EvAfoBe6RtD4iHmptzqxdNDuGBjowLJ+2h8UFnJ+vv23uu73tejD0cchq5RiyWjmGrFaOIetPLgp9wMlAd0Q8BiBpHTAXKFSAeqLuhhoRMdQO2vhOmWOoDx+zhmxExJDjoqFGRAwNRRufU1qlrWPIx5fGyUuh7yhgW9nrXmBW35UkLQWWppe/lvSzJuRtKI4Anml1Jvr6ywbkS1fW9PY31Skb5YYTQ7skPVLvjDRifxfxe2uNIUlLI+Kamj5lf8OKodmzZ/+KHP7f58lQY6TG2KhW3o9DuTyf1EuLjhv1lvcYqiTXcVXLeayJsVHPfdiOMVQPTY/Del4jDTHWmrGtVcVRXgp9qpAWByRkF3jXAEi6NyJmNjpjQ5HHPEF+81VnQ46hhmWkRft7JH4v9f0thxVDI+T/qyYjaB/V7ThU9H1W9O2rQUPPZXnf73nPH7RFHnNzPdSfVuzDkXaNVEleRu/sBY4uez0ZeLxFebH25BiyWjmGrFaOIauVY8hq5RiyivJS6LsHOFbSMZIOBs4D1rc4T9ZeHENWK8eQ1coxZLVyDFmtHENWUS6ad0bEHknLgA1kw8teFxEPDvK2llRJDyKPeYL85qtuhhlDjdKq/e3vrUENMVT4/686GBH7qM7HoaLvs6Jv37A04VyW9/2e9/xBzvOYs+uh/rRiHxbiWqUWijigma+ZmZmZmZkVRF6ad5qZmZmZmVkDuNBnZmZmZmZWYG1V6JN0rqQHJf1W0sw+yy6V1C3pEUlntjCPn5T0X5I2p8dZrcpLys+ctE+6Ja1oZV6KStLfSnpY0gOSviNpfD/r9UjakuLi3mF+14C/p6RDJN2Ylt8tacpwvqfPZx4t6U5JW9P/34UV1umU9HxZ3H+i1u9NnzvgPlPm6rS9D0iaXo/vrTJv/f6v5+V4lAc+Bg3NQMeTosSVY6K52iWm8hYX/Z37JE2QtFHSo+nv4a3Oaztq5vVyq2KrHtd9dRURbfMApgLHAV3AzLL044GfAocAxwA/B0a1KI+fBP7vVu+rlJdRaV+8GTg47aPjW52voj2AM4DR6fmVwJX9rNcDHNHI3xP4KPDV9Pw84MY6bF8HMD09Hwv8Z4Xv7QRubsC+HXCfAWcBt5HNS3QKcHcTf/eK/+t5Oh61+uFj0LD2WcXjSVHiyjHRkn2e+5jKY1z0d+4DPg+sSOkr+jvn+zHo/m3K9XIrY6vW6756P9qqpi8itkbEIxUWzQXWRcSvI+IXQDdwcnNzl0snA90R8VhE/AZYR7avrI4i4vsRsSe93EQ2J04jVPN7zgVWp+ffAk6XVGmi1qpFxBMRcX96vhPYChxVy2fW0VxgTWQ2AeMldeQgTz4eZXwMGqIBjidFiSvHRJO1SUzlLi4GOPeVn2dXA+e0JodWpdzFVqu0VaFvAEcB28pe99Lai9JlqRnFdS2u9s/bfhkJLiCreaokgO9Luk/S0mF8djW/57510kn+eeD1w/iuilJz0XcAd1dY/E5JP5V0m6QT6vSVg+2zVsd4pf/1VucpT7wvalN+PCnKvizKdrSrvMZUnvJygD7nvkkR8QRkBUNgYuty1vaacb3cytiq9bqvrnIxT185ST8A3lBh0WUR8d3+3lYhrWFzUQyUR+ArwKfT938aWEV2kG2Fpu6XIqsmLiVdBuwBvtHPx7wrIh6XNBHYKOnhiPj3oWSjQlrf37Nhv7mkMcC3gYsi4oU+i+8H3hQRu1K7/H8Fjq3D1w62zxoa48P8X/f/3au8LyoY5vGkKPuyKNuRKwWIqTzlZT99z301Np4ZUXJyvdzK2Kr1uq+uclfoi4h3D+NtvcDRZa8nA4/XJ0cHqjaPkv4RuLlR+ahCU/dLkQ32m0taBLwPOD1SQ+4Kn/F4+vuUpO+QNTkYyj9/Nb9naZ1eSaOB1wHPDuE7KpL0GrKT3jci4l/6Li8vBEbErZK+LOmIiHimlu+tYp81NMaH+b/u/7tXeV9UMMzjSVH2ZVG2I1cKEFN5yss+/Zz7npTUERFPpO4ET7Uuh/mWk+vllsVWHa776qoozTvXA+cpG7nwGLIahh+3IiN9+hO9H/hZK/KR3AMcK+kYSQeTDeyxvoX5KSRJc4BLgD+OiJf6WecwSWNLz8k61g81Nqr5PdcDi9LzPwHu6K8QWq3UJ/BaYGtE/F0/67yh1HdQ0slkx5Zf1fi91eyz9cBCZU4Bni81u2m0Af7Xc3M8ygEfg4ZogONJUeLKMdFkbRJTuYuLAc595efZRUB/rdBsAE28Xm5JbNXpuq+uclfTNxBJ7we+CBwJ3CJpc0ScGREPSroJeIis6cLHImJvi7L5eUknkVUd9wB/3qJ8EBF7JC0DNpCNXnRdRDzYqvwU2D+QjX62MZV7NkXEX0h6I/C1iDgLmAR8Jy0fDXwzIr43lC/p7/eU9Cng3ohYT3aCukFSN1kN33l12L53AecDWyRtTml/DfxuytdXyQqYH5G0B3gZOK/Wwib97DNJf1H2vbeSjeDZDbwEfLjG7xyKiv/rOTsetZSPQcNS8XhSlLhyTLRE7mMqp3HR37lvJXCTpCXAL4FzW5S/dteU6+UWxlbN1331ptqvy8zMzMzMzCyvitK808zMzMzMzCpwoa9JJG1IzfD6ps+VtF3SX0n6maSdkn4h6a/6rHenpKclvaBsWPwROcfISFZrDJWt/weSQtJnGp9ry5M6HId6JL0saVd6fL95ubc8qMdxSNKFadmLkrZKemtzcm95UEsMSfrdsuNP6RGSljd3K6yV6nAuO0nSf0h6XlKvpE80L/et40Jf81wPnJ86Bpc7n2z4ZAELgcOBOWRzl5T3x7oQ6IiIccBS4Otq/STU1lzXU1sMlUYiu4rK8+xZ8V1PjTEE/FFEjEmPMxqdYcud66khhiT9GbAEOBsYQzaiZE2j/FrbuZ5hxlBE/LLs+DMGmAb8lmyETRs5rqe2c9k3yUbRnAD8AdmYBH/c6Ey3mvv0NYmkQ4HtZBdM/57SDgeeAGZFxE/7rH812e/z8QqfVRry9bSIaMfR22wY6hFDklaQHeQmAr0R8TfNyr+1Xq0xJKkH+LOI+EFTM265UUsMSToI+P+AxRFxe5OzbjlR5+uhy4HOiJjd+JxbXtThXPYSMDMiHkqv/xm4PyKuaOJmNJ1r+pokIl4GbiK781DyAeDhCsEp4PeBB/uk3yzpFbJami7g3kbm2fKl1hiS9CayiU8PaBJhI0M9jkPAN5Q1Nf++pBMbmmHLnRpjaHJ6vE3SttTs6n+mwqCNEHU6DpUsBFY3Ip+WX3WIob8nm+7pNZKOA94JFP5mpg+0zbUaODfdoYD+D1afJPtt/qk8MSLeB4wlG6J+Q0T8tnFZtZyqJYauBv5HROxqaA4t72qJoQXAFOBNwJ3ABknjG5ZTy6vhxtDk9PcMsmZ5s4H5ZM09bWSp6XoIQNLvkw2L/60G5dHyrZYYuplsqqmXgYeBayPinsZlNR9c6GuiiPgh8DQwV9Kbgf9G1q54H2VziSwEzo6IX1f4jN0RcRtw5khof2z7G24MSfojYGxE3NjkLFvO1HIcioi7IuLliHgpNYPZQXYH1UaQGmLo5fT38xGxIyJ6gP9FdiPTRpB6XA+RTYz+bd/IHJlquB6aAHyPrNXT7wBHk11Tf7SJ2W+JtpqcvSDWkAXgccD3I+LJ0gJJFwAryPrq9Q7yOaOB32tYLi3PhhNDpwMzJW1Pr18H7JU0LSI8EuzIU6/jUJB1mLeRZzgx9AjwG7K4MRv2cSjV7pwLvL9JebV8Gk4MvRnYGxFr0uteSevIbj59uTnZbpGI8KOJD7KmUb8BeoFzy9IXkHVKnVrhPf8H8F7gUOA1wJ+mz5je6u3xo21iaCzwhrLHjcAXgAmt3h4/2iaGfhd4F3Aw2d3RvyK7y/r6Vm+PH+0RQ2n5GrKmVWPJmns+DCxp9fb40fzHcGMorfMhskGB1Ort8KN1j2Gey8aRtVL5EFmLxzcAPwI+2+rtafTDo3e2gKQu4ETgDfFqdfMvyE6A5U0Yvh4RfyFpKtnwtMcDe4FHgc9FxHeamW/Lj6HGUIX3X49H7xzRhnEcOgFYS9bC4BVgM3BJRHhAqRFqOMchSeOAa8imbNgB/CPw6fDFyIg03HOZpA3AjyPifzQxu5ZDwzwO/SFwJfBWsmbn/wZcGBEvNTHrTedCn5mZmZmZWYF5IBczMzMzM7MCc6HPzMzMzMyswFzoMzMzMzMzKzAX+szMzMzMzAqsbefpO+KII2LKlCn9Ln/xxRc57LDDmpehBiratjz88MPPRMSRrc7LYDFUksf9P9LzdN9997VVDJXk8XcbTFHz3K4x1ErtGAvQuHy3cwy1629Zb63eD+0YQ63eZ5XkLU/Nzk/VcdTqOSOG+5gxY0YM5M477xxweTsp2rYA90YbxFB5nvNmpOep3WKoJI+/22CKmud2jaFWasdYiGhcvts5htr1t6y3Vu+HdoyhVu+zSvKWp2bnp9o4cvNOMzMzMzOzAnOhz8zMzMzMrMBc6DMzMzMzMyswF/rMzMzMzMwKrG1H7yyyKStu2e/18ml7WNwnraRn5dnNyJLlRCk2BooJcFxYfUxZccugsQaOt6Loe+7pqxQL/r2tnZXHua+v2puviYbGNX1mZmZmZmYF5kKfmZmZmZlZgbnQZ2ZmZmZmVmAu9JmZmZmZmRWYC31mZmZmZmYF5kKfmZmZmZlZgbnQZ2ZmZmZmVmAu9JmZmZmZmRWYC31mlnuvvPIKJ598MieeeCInnHACwBsBJB0j6W5Jj0q6UdLBKf2Q9Lo7LZ9S+ixJl6b0RySdWZY+J6V1S1rR3C00MzMzaxwX+sws9w455BDuuOMOfvrTn7J582aAcZJOAa4EvhARxwLPAUvSW5YAz0XEW4AvpPWQdDxwHnACMAf4sqRRkkYBXwLeCxwPzE/rmpmZmbU9F/rMLPckMWbMGAB2794NICCAPwS+lVZbDZyTns9Nr0nLT5eklL4uIn4dEb8AuoGT06M7Ih6LiN8A69K6ZmZmZm1vdKszYGZWjb179zJjxgy6u7sBXgB+DuyIiD1plV7gqPT8KGAbQETskfQ88PqUvqnsY8vfs61P+qxK+ZC0FFgKMGnSJLq6uqrehl27dg1p/VZbPm0Pkw7N/g4kb9vUbvvZzMys0VzoM7O2MGrUKDZv3syOHTs4/PDDDwOmVlgt0l/1s6y/9EqtHqJCGhFxDXANwMyZM6Ozs3OwrO/T1dXFUNZvtcUrbmH5tD2s2jLwqaJnQWdzMlSldtvPZmZmjebmnWbWVsaPHw+wEzgFGC+pVCKZDDyenvcCRwOk5a8Dni1P7/Oe/tLNzMzM2p4LfWaWe08//TQ7duwA4OWXXwYYB2wF7gT+JK22CPhuer4+vSYtvyMiIqWfl0b3PAY4FvgxcA9wbBoN9GCywV7WN3q7zMzMzJrBzTvNLPeeeOIJFi1axN69e/ntb38L8EJE3CzpIWCdpM8APwGuTW+5FrhBUjdZDd95ABHxoKSbgIeAPcDHImIvgKRlwAZgFHBdRDzYxE00MzMzaxgX+sws997+9rfzk5/8ZN9rSU8ARMRjZCNv7iciXgHOrfRZEfFZ4LMV0m8Fbq1Tls3MzMxyw807zczMzMzMCsyFPjMzMzMzswJzoc/MzMzMzKzAXOgzMzMzMzMrMBf6rOG2bdvG7NmzmTp1KosXLwaYCCBpgqSNkh5Nfw9P6ZJ0taRuSQ9Iml76LEmL0vqPSlpUlj5D0pb0nqslVZqE28zMzMxsxHGhzxpu9OjRrFq1iq1bt/LlL38ZYKKk44EVwO0RcSxwe3oN8F6y+dOOBZYCX4GskAhcDswiG7Hx8lJBMa2ztOx9c5qxbWZmZmZmeedCnzVcR0cH06dnlXWvfe1rAV4GjgLmAqvTaquBc9LzucCayGwCxkvqAM4ENkbEsxHxHLARmJOWjYuIH6UJuNeUfZaZmZmZ2YjmefqsqbZv3w7wWuBuYFJElOZbe0LSxLTaUcC2srf1prSB0nsrpB9A0lKyGkEmTZpEV1fXoHnetWtXVes1w/JpewCYdOirzytpRX7ztJ/M+tq2bRsLFy5k+/btHHTQQVD/VbmiAAAgAElEQVTWzBy4EZgC9AAfiIjnUhPxq4CzgJeAxRFxf3rPIuBv0kd/JiJWp/QZwPXAoWRzPl6YbkSZmZm1VNWFPkmjgHuB/4qI90k6BlgHTADuB86PiN9IOoSspmUG8CvggxHRkz7jUmAJsBf4y4jYkNLnkJ1cRwFfi4iVddo+y5Fdu3bxiU98AmBbRLwwQLe7SgtiGOkHJkZcA1wDMHPmzOjs7Bwk11kBqpr1mmHxiluArMC3akv//749CzqblKNX5Wk/mfVVamY+ffp0du7cybhx40rNzBeTNTNfKWkFWTPzS9i/mfkssibks8qamc8kO87cJ2l9an1Qama+iazQNwe4rakbag23d+9eZs6cCfAWAF8P2VA5hqwVhtK880Jga9nrK4EvpP5Yz5EFHunvcxHxFuALaT3SyfU84ASyE+GXJY1KhckvkZ1gjwfmp3WtQHbv3s28efN497vfDbAjJT+ZmmaS/j6V0nuBo8vePhl4fJD0yRXSzcyA/ZuZjx07FtzM3IbpqquuYurUqeVJvh6yIXEMWStUVdMnaTJwNvBZ4OLU7OUPgQ+lVVYDnyS7yzk3PQf4FvAPaf25wLqI+DXwC0ndZINxAHRHxGPpu9aldR+qacssNyKCJUuWMHXqVM455xy+8pWvlBatBxYBK9Pf75alL0uxMAt4PjX/3AB8rmzwljOASyPiWUk7JZ1C1mx0IfDF5mydmbWbnp4eaLNm5s0wUJNxeLVZeV7yW616Nj1/+umn+frXv86f/umfsnbtWnw9ZEPV29vLLbfcwmWXXeYYsqaqtnnn3wP/HRibXr8e2BERpTNE+clt3wkxIvZI/z97dx9uV1Uf+v77u4lQhCAgJKaE02BJeUKIRkiBPnIwkQoRPcXegAfMlaBpsRZavXDuIdRzD15fCnpKqbTWHpSUkGuJtNVjLkSQirs9pQ3yIvIqTcR9TORNDGC2ghD83T/WWLDYWXvvtbNf1pxrfz/Ps5+95phjrj3mnL895xxzjjlGPFPyH0yjyQttlhl8Aj22XSFGc6Ks8/tFg0+8w72/VYd1vPfee1m3bh2vf/3r2bBhA8AREXEKjcredRGxCvgBcHpZZCON92i20HiX5n0ApXL3ceD2ku9jmbm9fP4gL79L8zVsUiWpjYGBAZYvXw41a2Y+GZrNx4fSbFbejebjYzGeTc9PO+00Pv/5z7Njx45mUuWvh9qp8zXSWLVeT3Xj+uriiy/mPe95D/fcc08zqVYxVKXYqWo/B1XaRq1GrPRFxDuBJzLzzohY0kxukzVHmDdUersmpmM+Udb5/aLBJ97h3t+qw8l3yZIl/MEf/AHQ2C9Lly59IDM3ltknDs5fmkad2+67MnMNsKZN+h3AkeNWaEk9p9nMfMWKFdx1112vaGZenvJ12sx8yaD0Pmxm3vOuv/56Zs6cydFHH916QVf566F26nyNNFat11iTfX11/fXXs2DBAj7wgQ/UNoaqFDtV7eegStuoVSfv9L0Z+K2I6KfxkulbaTz52y8imlu49eT20omyzH8NsJ3Rv6clSdK4aG1mfv7557fOajYzh12bmZ8VDcdRmpkDNwEnRcT+pan5ScBNZd6OiDiuNL86q+W71ANuvfVWNmzYwNy5cznjjDOg0frJ6yF1zBhSN41Y6cvMizJzTmbOpfHS6C2ZuQL4JnBayTb4RNk8gZ5W8mdJPyMi9iy9FM0DvkWjqd68iDg0IvYof2PDuKydJEk0LrbWrVvHLbfcwqJFi+CVzczfFhGbgbeVaWg0M3+YRjPzzwO/D41m5kCzmfnt7NrM/Atlme9hM/Oecskll7Bt2zb6+/tZv349wA6vhzQaxpC6aSzj9F0IrI+ITwDfBq4q6VcB68pLpdtpBByZeX9EXEfjZdKdwLmZ+SJARJxH4+7pNGBNZt4/hnJJkvQKxx9/PK1D5kWEzcw1Xrwe0lgZQ5pwo6r0ZWYfjXcXKD0DHdMmz3O83CHH4HmfpNED6OD0jTTuqkqSJFVaeV9nC3g9pN1jDGmyjWacPkmSJElSzVjpkyRJkqQeZqVPkiRJknqYlT5JkiRJ6mFW+iRV3tatW1m6dCnz589nwYIFADMBIuKAiLg5IjaX3/uX9IiIKyJiS0TcExFHNb8rIlaW/JsjYmVL+tERcW9Z5ooy1pokSVLtjWXIBkmaFNOnT+eyyy7jqKOOYseOHey7774zI+II4GzgG5l5aUSsBlbT6Pr67TTGLZoHHAt8Djg2Ig4ALgYWAwncGREbMvOpkuccYBONns+W0YPjrM1dfUO3iyBJkiaZT/okVd7s2bM56qjGw7oZM2YAPAscDJwKrC3Z1gLvKp9PBa7Jhk3AfhExGzgZuDkzt5eK3s3AsjJv38z81zI+2zUt3yVJklRrPumTVCv9/f0ArwZuA2Zl5qMAmfloRMws2Q4GtrYstq2kDZe+rU36LiLiHBpPBJk1axZ9fX0dl31gYGBU+SfCBQt3jir/rL1GXqbb6zRYFbazJElVYqVPUm0MDAywfPlygK2Z+ZNhXrtrNyN3I33XxMwrgSsBFi9enGWA3Y709fUxmvwT4exRNu+8YOFOLrt3+FNF/4olYyjR+KvCdpYkqUps3impFl544QWWL1/OihUrAJ4uyY+XppmU30+U9G3AIS2LzwEeGSF9Tpt0SZKk2rPSJ6nyMpNVq1Yxf/58zj///NZZG4BmD5wrga+2pJ9VevE8DnimNAO9CTgpIvYvPX2eBNxU5u2IiONKr51ntXyXJElSrdm8U1Ll3Xrrraxbt46FCxeyaNEigCMi4hTgUuC6iFgF/AA4vSyyETgF2AL8DHgfQGZuj4iPA7eXfB/LzO3l8weBq4G9aPTa2XM9d0qSpKnJSp9UAXajP7zjjz+eRqeaDRHxQGZuLJMnDs5feuA8t913ZeYaYE2b9DuAI8elwJIkSRVi805JkiRJ6mFW+iRJkiSph1npkyRJkqQeZqVPkiRJknqYlT5JkiRJ6mFW+iRJkiSphzlkgyRJ6thohpjpv/QdE1gSSVKnfNInSZIkST3MSp8kSZIk9TArfZIkSZLUw6z0SZIkSVIPs9InSZIkST3MSp8kSZIk9TArfZpw73//+5k5cyZHHnnkS2kRcUBE3BwRm8vv/Ut6RMQVEbElIu6JiKNalllZ8m+OiJUt6UdHxL1lmSsiIiZ1BSVJkqQKs9KnCXf22Wdz4403Dk5eDXwjM+cB3yjTAG8H5pWfc4DPQaOSCFwMHAscA1zcrCiWPOe0LLdswlZGUm15A0qSNFWNWOmLiEMi4psR8WBE3B8RHyrpnijVkRNOOIEDDjhgcPKpwNryeS3wrpb0a7JhE7BfRMwGTgZuzsztmfkUcDOwrMzbNzP/NTMTuKbluyTpJd6A0lhs3bqVpUuXMn/+fBYsWAAwE7weUueMIXXT9A7y7AQuyMy7ImIGcGdE3AycTeNEeWlErKZxoryQV54oj6VxEjy25US5GMjyPRvKBXzzRLkJ2EjjRPm18VtNVdCszHwUIDMfjYiZJf1gYGtLvm0lbbj0bW3S24qIc2jEGrNmzaKvr2/Egg4MDHSUbywuWLhzVPln7TX8MhNd3nYmYztJY3HCCSfQ398/OPlUYEn5vBboo3Eue+kGFLApIpo3oJZQbkABlPPhsojoo9yAKunNG1Cey3rE9OnTueyyyzjqqKPYsWMH++6778yIOAKvh9QhY0jdNGKlr1yYNy/Od0TEgzQuqj1RaiK0uyOVu5HeVmZeCVwJsHjx4lyyZMmIBerr66OTfGNx9uobRpX/goU7uezeof99+1csGWOJRm8ytpM0AbpyA0r1M3v2bGbPng3AjBkzAJ7F6yGNgjGkburkSd9LImIu8CbgNrpwohzNU5o6P3UY/ARnuKc6dVnHxx57jJ/+9KcMDAw0kx6PiNkldmYDT5T0bcAhLYvOAR4p6UsGpfeV9Dlt8kvSWEzYDajdaXEwFvf+8JmO8l2wcPj5I7UwaKcK56iJuB547LHHAF5NDa6H2qnzNdJYtcZwN6+v6hpDVYqd5r6rWuunKm2jVh1X+iJiH+DvgQ9n5k+GaSI8YSfK0TylqfNTh8FPfYZ7qtONJzq7o7+/n7333pt99tmnmbQBWAlcWn5/tSX9vIhYT6MpwzPlAHgT8Mct786cBFyUmdsjYkdEHEfjwHkW8OeTs1bVNXcUTw77L33HBJZEqrxJvwG1Oy0OxmK0LQmGMlILg3aqcI4a7+uBgYEB3vKWtwBsrcP1UDt1vkYaq9b/h25dX9U5hqoUO819WbXWT1XaRq066r0zIl5Fo8L3xcz8ckl+vJwgGcWJcqh0n9T0sDPPPJPf+I3f4KGHHuL0008HOJBGZe9tEbEZeFuZhkb784eBLcDngd8HKE0YPg7cXn4+1mzWAHwQ+EJZ5nvYjEFS55o3oGDXG1BnlY4UjqPcgAJuAk6KiP3LTaiTgJvKvB0RcVzpOOGslu9Sj3jhhRdYvnw5K1asAHi6JHs9pI4ZQ+qWTnrvDOAq4MHM/NOWWZ4o1ZFrr72WRx99lBdeeIG//du/BXgyM3+cmSdm5rzyeztA6bXz3Mz81cxcmJl3NL8nM9dk5mHl569b0u/IzCPLMueVtu/qIXa1r/HQegMKeENErMIbUOpQZrJq1Srmz5/P+eef3zrL6yF1xBhSN3XSTuPNwHuBeyPi7pL2RzROjNeVk+YPgNPLvI3AKTROej8D3geNE2VENE+UsOuJ8mpgLxonSU+Ukl5y9tlnc95553HWWWe1Jje72re3M3Xk2muvfelzRNyTmVeVyRMH5y03j85t9z2ZuQZY0yb9DuDIXZdQL7j11ltZt24dCxcuZNGiRQBHRMQpeD2kDhlD6qZOeu/8Z9q3EQZPlJImgV3tS+q2448/ntaGJBHxQGZuLJNeD2lExpC6qaN3+iSpgl7R2xllkFvsal+SJOkVRtcNlyRV34SO9TiWrtKr0I3zaLvc76Sb/m6v02BV2M6SJFWJlT5JddWVsR7H0lV6FbpxHm33/Z1001+FbvlbVWE7S5JUJTbvlFRX9nYmSZLUAZ/0Saq8M888k76+Pp588knmzJkDL4/1aG9nkiRJI7DSJ6nyWrvaB4iIJzPzx9jbmSRJ0ohs3ilJkiRJPcxKnyRJkiT1MCt9kiRJktTDfKdPmkBzR9k9viRJkjTefNInSZIkST3MSp8kSZIk9TCbd0qSpAnRaRP3/kvfMcElkaSpzSd9kiRJktTDrPRJkiRJUg+z0idJkiRJPcxKnyRJkiT1MCt9kiRJktTDrPRJkiRJUg+z0idJkiRJPcxKnyRJkiT1MCt9kiRJktTDrPRJkiRJUg+z0idJkiRJPcxKnyRJkiT1MCt9kiRJktTDrPRJkiRJUg+b3u0CSHU0d/UN3S6CJEmS1JHKVPoiYhnwGWAa8IXMvLTLRVLNGEMaK2No93R6E6T/0ndMcEm6bzJjyJtPvcnjkMbKGFI7laj0RcQ04LPA24BtwO0RsSEzH+huyVQXxpDGqu4xZAWg++oeQ+o+Y0hjVfcY8lw2cSpR6QOOAbZk5sMAEbEeOBWoRYB2ykCeUFMihjShjCGNlTG0m3xa/BJjaDd5jfUSY0htVaXSdzCwtWV6G3Ds4EwRcQ5wTpkciIiHhvnOA4Enx62EXfSHw6xLfGqSCzN2BwK/MgHfOxEx1FS5WBouJkZrHGNoMrdT3WKoqXKxNJKKxtpIOilzXWOoa8YzFnbXbsbQRJW7zjHU9X1ZBRW4vqpjDFUudkY6NnXhWnmyt1FHcVSVSl+0SctdEjKvBK7s6Asj7sjMxWMtWBX04LrMnYivbpM2phh66YsruP0t04SYsBh66Q/UcBtZ5tH96TZp4xpD3VTHWIDalXtSYqhm22TC9Oh2mNAYquI2q1qZqlaepqoM2bANOKRleg7wSJfKonoyhjRWxpDGyhjSWBlDGitjSG1VpdJ3OzAvIg6NiD2AM4ANXS6T6sUY0lgZQxorY0hjZQxprIwhtVWJ5p2ZuTMizgNuotG97JrMvH+MX1u7pjPDcF1GMEEx1FTF7W+ZxtkEx1BTHbeRZe7QJMVQN9UxFqBG5Z7EGKrNNplgPbcdJiGGqrjNqlamqpUHgMjcpZmvJEmSJKlHVKV5pyRJkiRpAljpkyRJkqQe1nOVvoj4bxHx3Yi4JyK+EhH7tcy7KCK2RMRDEXFyN8vZqYhYVsq7JSJWd7s8oxERh0TENyPiwYi4PyI+VNIPiIibI2Jz+b1/t8vaNFz8DMrXHxH3RsTdEXHHBJVl2H0fEXtGxJfK/NsiYu5ElKPl77Xdn4PyLImIZ8p2uTsi/utElqmOIuKjEfHDlm10SrfL1E5djz2T8b/Z66p27BmJx6bO9do10ljU9RjXLRFxevn/+kVELB40ryuxU4V9GBFrIuKJiLivJa2a17mZ2VM/wEnA9PL5U8CnyucjgO8AewKHAt8DpnW7vCOsy7RSztcDe5TyH9Htco2i/LOBo8rnGcC/lf3waWB1SV/d3EdV+Bkqftrk6wcO7Oa+B34f+Kvy+QzgS93Yn4PyLAGu7/Z+rPIP8FHgP3W7HCOUsbbHnon+3+z1nyoeezoos8emzrdVz1wjjXE71PYY18VtNh84HOgDFrekdyV2qrIPgROAo4D7WtIqeZ3bc0/6MvPrmbmzTG6iMT4JwKnA+sz8eWZ+H9gCHNONMo7CMcCWzHw4M58H1tNYj1rIzEcz867yeQfwIHAwjXVYW7KtBd7VnRLuapj4mWyd7PvW7fh3wIkR0W5Q1nExzP5U76n1sUdjUrljz0g8NnWux66RxsJj3Chl5oOZ+VCbWd2KnUrsw8z8J2D7oORKXuf2XKVvkPcDXyufDwa2tszbRvVPCnUsc1ul+c+bgNuAWZn5KDRO1sDM7pVsWK3xM1gCX4+IOyPinAn4253s+5fylJP4M8BrJ6Asuxi0Pwf7jYj4TkR8LSIWTEZ5aui80rxqTWWafbxSnY89E/2/2esqfewZicemUan7NdJYTLX1nUjd2pZV3oeVvM6txDh9oxUR/wC8rs2sj2TmV0uejwA7gS82F2uTv+rjVdSxzLuIiH2Avwc+nJk/6eIN4WZ5did+BntzZj4SETOBmyPiu+Vuz7gVs03a4H3flfgYvD8Hzb4L+JXMHCjvqv0PYN5El6lqhosx4HPAx2nsq48Dl9G4+KqSOh97Jvp/s9dV9tgzEo9NDVPoGmksptr6dqST2Gm3WJu0ydiW7sNRqmWlLzN/c7j5EbESeCdwYpYGtTTuABzSkm0O8MjElHDc1LHMrxARr6JxEv5iZn65JD8eEbMz89GImA08MZll2s34Gfwdj5TfT0TEV2g0MxjPC8tO9n0zz7aImA68hl2bGIyrIfbnS1ovtDJzY0T8ZUQcmJlPTmS5qmakGGuKiM8D109wcXZHbY89k/C/2esqeewZiceml02ha6SxmGrr25FOz12DdGtbVnkfdvU6dyg917wzIpYBFwK/lZk/a5m1ATij9Dp2KI07fN/qRhlH4XZgXkQcGhF70HhhfkOXy9Sx8o7HVcCDmfmnLbM2ACvL55XAUHePJt0w8dOaZ++ImNH8TOPF+Pva5R2DTvZ963Y8DbhlqErqeBhmf7bmeV3z3Z6IOIbGMebHE1WmOiongKbfZvxjZzzU8tgzSf+bva5yx56ReGzqXI9dI41FLY9xFdWt2KnyPqzkdW4tn/SN4C9o9CB0czm+b8rM38vM+yPiOuABGk0azs3MF7tYzhFl5s6IOA+4iUYvRWsy8/4uF2s03gy8F7g3Iu4uaX8EXApcFxGrgB8Ap3epfO20jZ+I+GXgC5l5CjAL+EqZPx34m8y8cTwLMdS+j4iPAXdk5gYaFznrImILjbvsZ4xnGdoYan/+u1Lmv6JxAfjBiNgJPAuc0c2LwYr6dEQsotEMpR/4QHeLs6saH3sm/H+z11X02DMSj02d65lrpLGo8TGuayLit4E/Bw4CboiIuzPz5G7FTlX2YURcS6N34AMjYhtwMRW9zo2pecyTJEmSpKmh55p3SpIkSZJeZqVPkiRJknqYlb4uiYibyjsSg9NPjYjHIuI/RcTDEfGTiHgkIi4vvaRJQEcxNL1M7xER3y1tzaWXdBBDn4iIFyJioOXn9d0oq6qpk+NQRBwVEf9U4ufxiPhQN8qqauoghr426Bj0fETc242yqpo6iKE9I+KvyvFne0T8fxFRlTH9Jo2Vvu65GnhvszexFu+lMW7OV4CjMnNf4EjgjcAfTmoJVXVXM0wMlUGTAf4vKtJdsCrnaoY/Du0EvpSZ+7T8PDzZhVSlXc3wMbQfcCPw32kM4H4Y8PXJLKAq72qGP5e9vfUYBPwL8LeTXUhV2tUMfxz6EPAbwBuAXwaeptEpzZRipa97/gdwAPDvmwkRsT+NsXOuyczvZebTzVnAL2icLKWmYWOoTB8K/B/AJd0ooCpvxBiSRjBSDJ0P3JSZX8zMn2fmjsx8sDtFVUV1fByKiLkl37rJK55qYKQYOpTGcejxzHwOWA8s6EZBu8lKX5dk5rPAdcBZLcnvBr6bmd8BiIj3RMRPgCdpPOn775NeUFVWJzFE407WH9Hoolx6hQ5j6D+U5jD3R8QHJ72QqrQOYug4YHtE/EtEPFGaVf27bpRV1dThcajpLOB/Zub3J6t8qr4OYugq4M0R8csR8WpgBfC1yS9pd1np6661wOkRsVeZPqukAZCZf1Oad/4a8FfA45NfRFXckDFUxtSZnplf6VbhVAvDHYeuA+bTGJfpd4H/GhFnTn4RVXHDxdAcGoMTf4jGuHnfB66d9BKq6oa9HmpxFo2mfNJgw8XQv9EYL++HwE9onNd2eQew1zlOX5eVwW3/C/At4LvAIZm5S+UuIs4A3p2Z//skF1EV1y6GgAHgbuCUzNwcEUuA/zcz53StoKqsURyHVgO/npnLJ7mIqrihYigivgPclZnvK/leS6P1yn6Z+UzXCqzKGek4FBHH03g/9HWZOdCdUqrKhjkOfRF4NbAK+Cnwn4F3ZuaxXStsF9gbZPddQ+NuxOHA19tdaBXTgV+dtFKpTnaJoYhYBMwF/md5r3kP4DUR8RhwXGb2d6msqqZOj0NJ4x1jabChYugeGnHT1PxsHGmwkY5DK4EvW+HTMIaKoTcCH8nM7QAR8efAxyLiwMx8sjtFnXw27+y+a4DfpNF06qWmDBHxOxExs3w+ArgI+EZXSqiqaxdD99F44reo/PwOjebBi4CtXSijqm2o49CpEbF/NBxDowfhr3apjKq2tjEE/DXw2xGxKCJeBfzfwD+3dFQmNQ0VQ5Qme6dj004Nb6gYuh04KyJeU45Dvw88MpUqfGClr+vKE5d/AfYGNrTMejNwb0T8FNhYfv5o0guoymsXQ5m5MzMfa/4A24FflOkXu1daVdEwx6EzgC3ADhon009lZrv3bDTFDRVDmXkLjXPXDTSGjjkMeE8XiqiKG+Y4BPAu4Bngm5NcLNXIMDH0n4DngM3Aj4BTgN+e7PJ1m+/0SZIkSVIP80mfJEmSJPUwK32SJEmS1MOs9EmqjYiYFhHfjojry/ShEXFbRGyOiC9FxB4lfc8yvaXMn9vyHReV9Ici4uSW9GUlbUsZmkCSJKknWOmTVCcfAh5smf4UcHlmzgOeojEGD+X3U5l5GHB5ydfsCfcMYAGwDPjLUpGcBnwWeDtwBHBmyStJklR7tR2n78ADD8y5c+d25W//9Kc/Ze+99+7K3x5v3ViXO++888nMPGhS/2gb3Yyh4fRSfA1ld9bx+eefZ8aMGbzuda9j8+bNz0djAMK38nJPgGuBjwKfA04tnwH+DviLkv9UYH1m/hz4fhnI9ZiSb0tmPgwQEetL3geGK9N+++2Xhx122KjWo9vqGF8TUea6H4fquB+HUtd1qXsMjacq7sM6lKnqMVTFbVg1VdhGncZRbSt9c+fO5Y477ujK3+7r62PJkiVd+dvjrRvrEhH/a1L/4BC6GUPD6aX4GsrurONpp53GX//1X7Njxw6WLl36LPBa4OnM3FmybAMOLp8PpoxHmJk7I+KZkv9gYFPL17Yus3VQ+rHtyhER5wDnABx00EH8yZ/8yajWo9sGBgbYZ599ul2MUZmIMi9durTWx6FeOk7UdV08l72sivuwDmWqegxVcRtWTRW2UadxVNtKn6Sp4/rrr2fmzJkcffTR9PX1NZOjTdYcYd5Q6e2aurcdzyYzrwSuBDj88MOz2wf70arCCWq06lhmSZKqxEqfpMq79dZb2bBhAxs3buS5554DmAH8GbBfREwvT/vmAI+URbYBhwDbImI68BoaA9Q305talxkqXZIkqdbsyEVS5V1yySVs27aN/v5+1q9fD7AjM1cA3wROK9lWAl8tnzeUacr8WzIzS/oZpXfPQ4F5wLeA24F5pTfQPWh09rJhMtZNkiRpovmkb4qYu/qGtukXLNzJ2S3z+i99x2QVSS1a98/gfdLK/bOLC4H1EfEJ4NvAVSX9KmBd6ahlO41KHJl5f0RcR6ODlp3AuZn5IkBEnAfcBEwD1mTm/ZO6JmM01P/4YFcv86V8TZ5O4vKChTtZMvFFUY8bHGtDnUs9j2qs6nrNZqVPUq2Ud7u2AJTeNo8ZnCcznwNOb7d8Zn4S+GSb9I3AxnEsqiRJUiXYvFOSJEmSepiVPknSlPHiiy/ypje9CeAwgPIe520RsTkivlTe6aS89/mliNhS5s9tfkdEXFTSH4qIk1vSl5W0LRGxenLXTJKkoVnpkyRNGZ/5zGeYP39+a9KngMszcx7wFLCqpK8CnsrMw4DLSz4i4gga74guAJYBfxkR0yJiGvBZ4O3AEcCZJa8kSV1npU+SNCVs27aNG264gd/5nd8BICICeCvwdyXLWuBd5fOpZZoy/8SS/1RgfWb+PDO/T+P90mPKz5bMfDgznwfWl7ySJHWdHblIkqaED3/4w3z60zVkBUoAACAASURBVJ9mx44dzaTXAk+XcR6hMY7jweXzwcBWgMzcGRHPlPwHA5tavrZ1ma2D0o9tV46IOAc4B2DWrFn09fWNel0GBgZ2a7nJdsHCnSPmmbUXtVgXSaozK32SpJ53/fXXM3PmTI4++ujWCka0yZojzBsqvV3LmWyTRmZeCVwJsHjx4iw90o5KX18fu7PcZBuqK/NWFyzcybtrsC6SVGdW+iRJPe/WW29lw4YNbNy4keeeew5gBvBnwH4RMb087ZsDPFIW2QYcAmyLiOnAa2iM+dhMb2pdZqh0SZK6ynf6JEk975JLLmHbtm309/ezfv16gB2ZuQL4JnBaybYS+Gr5vKFMU+bfkplZ0s8ovXseCswDvgXcDswrvYHuQaOzlw2TsW6SJI3EJ32SpKnsQmB9RHwC+DZwVUm/ClgXEVtoPOE7AyAz74+I64AHgJ3AuZn5IkBEnAfcBEwD1mTm/ZO6JpIkDaHjSl/pjvoO4IeZ+c5yh3M9cABwF/DezHw+IvYErgGOBn4M/MfM7C/fcRGNbrBfBP4wM28q6cuAz9A4UX4hMy8dp/WTJOkVyrtwWwAy82EaPW++QmY+B5zebvnM/CTwyTbpG4GN41hUSZLGxWiad34IeLBl2rGNJEmSJKniOqr0RcQc4B3AF8q0YxtJkiRJUg102rzzz4D/TKO3M6jx2EbjoS7jI7UaaqykWXu9cl7d1kuSJEnS8Eas9EXEO4EnMvPOiFjSTG6TtRZjG42HuoyP1GqosZIuWLiTy+59OQz6VyyZpBJJkiRJmgydPOl7M/BbEXEK8EvAvji2kSRJkiTVwojv9GXmRZk5JzPn0uiI5RbHNpIkSZKkehjLOH2ObSRJkiRJFTeqSl9m9gF95bNjG0mSJElSxY1mnD5J6ornnnuOY445hje+8Y0sWLAA4JcBSrPw2yJic0R8qTQRpzQj/1JEbCnz5za/KyIuKukPRcTJLenLStqWiFg9uWsoSZI0caz0Saq8Pffck1tuuYXvfOc73H333QD7RsRxwKeAyzNzHvAUsKossgp4KjMPAy4v+YiII2g0OV8ALAP+MiKmRcQ04LPA24EjgDNLXkmSpNqz0iep8iKCffbZB4AXXngBGkPAJPBW4O9KtrXAu8rnU8s0Zf6JERElfX1m/jwzvw9sodFM/RhgS2Y+nJnPA+tLXkmSpNobS0cukjRpXnzxRY4++mi2bNkC8BPge8DTZdgYaAwLc3D5fDCwFSAzd0bEM8BrS/qmlq9tXWbroPRj25UjIs4BzgE46KCD6OvrG9uKjZMLFu4cORMwMDBQmTJ3qo5lliSpSqz0SaqFadOmcffdd/P000+z//777w3Mb5Mty+8YYt5Q6e1aPWSbNDLzSuBKgMMPPzyXLFkyUtEnxdmrb+go39XL9qYqZe5UX19f7cosSVKV2LxTE27r1q0sXbqU+fPnNzvhmAkQEQdExM2lE46bI2L/kh4RcUXpUOOeiDiq+V0RsbLk3xwRK1vSj46Ie8syV5SmfOpB++23H8AO4Dhgv4ho3ryaAzxSPm8DDgEo819DYwiZl9IHLTNUuiRJUu1Z6dOEmz59OpdddhkPPvggmzZtAphZOslYDXyjdMLxjTINjc405pWfc4DPQaOSCFxMo9ndMcDFzYpiyXNOy3LLJmPdNDl+9KMf8fTTTwPw7LPPAuwLPAh8EzitZFsJfLV83lCmKfNvycws6WeU3j0PpREr3wJuB+aV3kD3oNHZy4aJXi9J9fPiiy/ypje9CeAwsBdhjZ4xpG6w0qcJN3v2bI46qvGwbsaMGQDP0niPqrWzjcGdcFyTDZtoPM2ZDZwM3JyZ2zPzKeBmYFmZt29m/mu5sL+m5bvUAx599FGWLl3KG97wBn79138d4CeZeT1wIXB+RGyh8c7eVWWRq4DXlvTzKTcUMvN+4DrgAeBG4NzMfLG8F3gecBONyuR1Ja96hMN+aLx85jOfYf78V7QutxdhjYoxpG7wnT5Nqv7+foBXA7cBszLzUYDMfDQiZpZsL3XCUTQ72xgufVub9F20dsIxa9asynQO0doJx6y9hu6UoyrlHavd6Zjj8ssvf+nz0qVLm3HzMI2nvq+Qmc8Bp7f7nsz8JPDJNukbgY2jKpRqoznsxz777MMLL7zAHnvs0Rz243waF1vrI+KvaFxkfY6Wi62IOIPGxdZ/HHSx9cvAP0TEr5U/81ngbTSOQbdHxIbMfGBy11QTadu2bdxwww185CMf4dprr6W8SvBW4D0ly1rgozRi6NTyGRq9CP/F4F6Ege+Xm1PN49iWclwjIpq9CBtDPcQYUrdY6dOkGRgYYPny5QBbM/Mnw7x2N9pOOIZK3zWxpROOxYsXV7ITjgsW7uSye9v/a/avWDJJJZpYdsyhyTbCsB9ebKkjH/7wh/n0pz/Njh07mkmvpcu9CFfhBmYVetgdfLN0qBuo3SznwMAAK1as4D3veQ/33HNPM7myMVSF/VpFdb1Rb6VPk+KFF15g+fLlrFixgrvuuuvpkvx4RMwuT/lmA0+U9OE621gyKL2vpM9pk7/r5nbYo6KkiVeVYT9UT9dffz0zZ87k6KOPbr2YG+6m46T0IlyFG5hVuJE3uAfjoW6gdvPm6SWXXMKCBQv4wAc+UIsYqsJ+raK63qi30qcJl5msWrWK+fPnc/7553PBBRc0ZzU727iUXTvhOK/cKT8WeKZUDG8C/ril85aTgIsyc3tE7ChNtW4DzgL+fHLWTlJdVGXYj/F4SlOXO/CdjB85a6/q3RFv59prr+XrX/86X/7yl3n++ecBZgB/RulFuNw8aNeL8LYOexFmmHT1gPvuu49//Md/ZOPGjTz33HNgDGkSWenThLv11ltZt24dCxcuZNGiRQBHRMQpNCp710XEKuAHvPwO1kbgFGAL8DPgfQClcvdxGj0tAnwsM7eXzx8Ergb2Ar5WfiRpF+2G/ZjMi63xeEpTlzvwnYwfecHCnby7BuvSur37+vpYunTpjsxcERF/S6OX4PW070X4X2npRTgiNgB/ExF/SuO90GYvwkHpRRj4IY13R9+Desbv/u7v8sUvfhEwhjT5Rqz0RcQvAf8E7Fny/11mXlwCaj1wAHAX8N7MfD4i9qTRe+LRwI+B/5iZ/eW7LqLxcvyLwB9m5k0lfRnwGWAa8IXMvHRc11Jddfzxx9PoVLMhIh4onWYAnDg4f+mB89x235WZa4A1bdLvAI4clwJL6jk/+tGPeNWrXsV+++031LAfXmxpd10IrI+ITwDf5pW9CK8r731upxETZOb9EdHsRXgnpRdhgIho9iI8DVhjL8JThjGkCdfJk76fA2/NzIGIeBXwzxHxNezxTJJUE48++igrV67kxRdf5Be/+AWUYT8i4gG82NIolad+W8BehLV7jCFNthErfeWpy0CZfFX5scczSVJtvOENb+Db3/72S9MR4bAfkqQpo6PB2cuAj3fT6F3xZkbR4xnQ2uPZaMZekyRJkiSNUUcduZSmK4siYj/gK9S4x7PxUJde01oN1YPa4PFF6rZekiRJkoY3qt47M/PpiOijxj2ejYe69JrWaqge1AaPL1K1MUUkSZIkjc2IzTsj4qDyhI+I2Av4TV7Z4xm07/EMWno8K+lnRMSepXezZo9nt1N6PIuIPWi8LL9hPFZOkiRJkqa6Tp70zQbWRsQ0GpXE6+zxTJIkSZLqoZPeO+8B3tQm3R7PJEmSJKniOuq9U5IkSZJUT1b6JEmSJKmHWemTJEmSpB5mpU+SJEmSepiVPkmSJEnqYVb6JFXe1q1bWbp0KfPnz2fBggUAMwEi4oCIuDkiNpff+5f0iIgrImJLRNwTEUc1vysiVpb8myNiZUv60RFxb1nmioiIyV5PSZKkiWClT1LlTZ8+ncsuu4wHH3yQTZs2AcyMiCOA1cA3MnMe8I0yDfB2YF75OQf4HDQqicDFwLE0hpy5uFlRLHnOaVlu2WSsmyRJ0kSz0iep8mbPns1RRzUe1s2YMQPgWeBg4FRgbcm2FnhX+XwqcE02bAL2i4jZwMnAzZm5PTOfAm4GlpV5+2bmv2ZmAte0fJckSVKtjTg4uyRVSX9/P8CrgduAWZn5KEBmPhoRM0u2g4GtLYttK2nDpW9rk76LiDiHxhNBDjroIPr6+sa0PuPlgoU7O8o3MDBQmTJ3qo5lliSpSqz0SaqNgYEBli9fDrA1M38yzGt37WbkbqTvmph5JXAlwOGHH55LliwZodST4+zVN3SU7+ple1OVMneqr6+vdmWWJKlKbN4pqRZeeOEFli9fzooVKwCeLsmPl6aZlN9PlPRtwCEti88BHhkhfU6bdPUIOwOSJE1lVvokVV5msmrVKubPn8/555/fOmsD0LzoXgl8tSX9rHLhfhzwTGkGehNwUkTsXy7uTwJuKvN2RMRx5UL9rJbvUg+wMyBJ0lRm805JlXfrrbeybt06Fi5cyKJFiwCOiIhTgEuB6yJiFfAD4PSyyEbgFGAL8DPgfQCZuT0iPg7cXvJ9LDO3l88fBK4G9gK+Vn7UI2bPns3s2bOBtp0BLSnZ1gJ9wIW0dAYEbIqIZmdASyidAQFERLMzoD5KZ0AlvdkZkHEkSeo6K32SKu/444+nce3dEBEPZObGMnni4PzlQv3cdt+VmWuANW3S7wCOHJcCV9i9P3ym4/f/+i99xwSXpjuq1BnQrFmzdquTmrp0btNJB0Oz9qIW6yJJdTZipS8iDqHRffnrgF8AV2bmZ0oTly8Bc4F+4N2Z+VRpGvUZGnfZfwacnZl3le9aCfyX8tWfyMy1Jf1oXr7DvhH4ULZe4UmSNA6q1hnQ4sWLd6szoLp0btPJDYYLFu7k3TVYF0mqs07e6dsJXJCZ84HjgHN9D0KSVDd2BiRJmqpGrPRl5qPNJ3WZuQN4EAdFliTViJ0BSZKmslG90xcRc4E3UeP3IMZDXd6laDXUexWz9nrlvLqtlyR1ws6AJElTWceVvojYB/h74MN1fg9iPNTlXYpWQ71XccHCnVx278th0L9iySSVSJImj50BSZKmso7G6YuIV9Go8H0xM79ckn0PQpIkSZIqbsRKX3k34Srgwcz805ZZvgchSZIkSRXXyZO+NwPvBd4aEXeXn+Z7EG+LiM3A28o0NN6DeJjGexCfB34fGu9BAM33IG5n1/cgvlCW+R6+B9FT3v/+9zNz5kyOPPLlVk8RcUBE3BwRm8vv/Ut6RMQVEbElIu6JiKNalllZ8m8uw38004+OiHvLMlfEMG2PJUnaHVu3bmXp0qXMnz+fBQsWAMwEz2fq3BNPPGEMqWs66b3znzMzMvMNmbmo/GzMzB9n5omZOa/83l7yZ2aem5m/mpkLyzsOze9ak5mHlZ+/bkm/IzOPLMuc5xh9veXss8/mxhtvHJzskB+SpNqYPn06l112GQ8++CCbNm0CmOkQVhqNadOmGUPqmo7e6ZPG4oQTTuCAAw4YnOyQH5Kk2pg9ezZHHdV40DJjxgyAZ3EIK43Ca1/7WmNIXTOqIRukcTTpQ37A5A/7MdRQGcMZPIxGq14ZUqOOw55IUlN/fz/Aq5niQ1g1VeGYPvi8OdS5tJvlbN1Ojz32GFQ8hqqwX6uoNa7qdM1mpU9VM2FDfsDkD/sx1FAZwxk8jEarXhlSo47DnkgSNC6Ely9fDrB1qg9h1VSFY/rg8+1Q59Junkeb22lgYIC3vOUtUPEYqsJ+raLWWKvTNZvNO9UtDvkhSaqVF154geXLl7NixQqAp0uy5zN1zBhSt1jpU7c45IckqTYyk1WrVjF//nzOP//81lmez9QRY0jdZPNOTbgzzzyTvr4+nnzySebMmQNwII0hPq6LiFXAD4DTS/aNwCk0hu/4GfA+aAz5ERHNIT9g1yE/rgb2ojHch0N+SFKNzO2wKXz/pe+Y4JIM7dZbb2XdunUsXLiQRYsWARwRLw9h5flMI7rvvvuMIXWNlT5NuGuvvfYV0xHxZGb+GDhxcN7S29S57b4nM9cAa9qk3wEcuesSkiSNj+OPP57WEaUi4oHM3FgmPZ9pRAsXLjSG1DU275QkSZKkHmalT1Llvf/972fmzJkceeTLNy8j4oCIuDkiNpff+5f0iIgrImJLRNwTEUe1LLOy5N8cEStb0o+OiHvLMlfEMF2pSZIk1Y2VPkmVd/bZZ3PjjTcOTl4NfCMz5wHfKNMAbwfmlZ9zgM9Bo5IIXAwcCxwDXNysKJY857Qst2zCVkaSJGmSWemTVHknnHACBxxwwODkU4G15fNa4F0t6ddkwyZgv9IF9snAzZm5PTOfAm4GlpV5+2bmv5b3J65p+S5JkqTas9Inqa5mle6pKb9nlvSDga0t+baVtOHSt7VJV4+xmbAkaaqy905JvabdhXbuRnr7L484h0ZTUA466CD6+vp2o4jj74KFOzvKN2uvzvNWZd0GBgbGpSxvfOMbOe6447jkkktak5vNhC+NiNVl+kJe2Uz4WBpNgI9taSa8mEac3BkRG8rT42Yz4U00ulpfht2lS5IqYMRKX0SsAd4JPJGZR5a0A4AvAXOBfuDdmflUuav5GRpjivwMODsz7yrLrAT+S/naT2Tm2pJ+NC+PJ7IR+FC29mcrSe09HhGzM/PR0kTziZK+DTikJd8c4JGSvmRQel9Jn9Mmf1uZeSVwJcDhhx+eS5YsGSrrpDq7w3HOLli4k8vu7ex+X/+KJWMo0fjp6+tjPLbzkiVL6O/v54orrmhNPpWX42ItjZi4kJZmwsCmiGg2E15CaSYMEBHNZsJ9lGbCJb3ZTNhKnySp6zo5818N/AWN91yavDMqqds2ACtpDGq7EvhqS/p5EbGexnHomVIxvAn445bOW04CLiqD3O6IiOOA24CzgD+fzBVRV72imXBETHgz4danxbNmzdqtp5jj9fRzonXyVLmOT58lqW5GrPRl5j9FxNxByd4ZlTRpzjzzTPr6+njyySeZM2cOwIE0KnvXRcQq4AfA6SX7RhqtDbbQaHHwPoBSufs4cHvJ97HmMQn4IC+3OPgaHoM0gc2EW58WL168eLeeFo/X08+J1skT6Do+fZakutndd/om/c4ojM/d0fFQlzusrYa6izr4Dmvd1ktTw7XXXvuK6Yh4MjN/DJw4OG+56XRuu+/JzDXAmjbpdwBH7rqEpoCuNBOWJGkyjXdHLhPagcJ43B0dD3W5w9pqqLutg++wehdV0hRjM2FJUs/b3Uqfd0YlSbXS2kwYeENpGmwzYUlSz9vdSp93RqUumNthD439l75jgksi1U9rM+GIuCczryqTNhOWJPW0ToZsuJbGU7oDI2IbjV44vTMqSZIkSTXQSe+dZw4xyzujPcgnSZIkSVJv+d+6XQBJkiRJ0sSx0idJkiRJPcxKnyRJkiT1MCt9kiRJktTDrPRJkiRJUg+z0idJkiRJPWx3B2eXprROh7aQJEmSus0nfZIkSZLUw6z0SZIkSVIPs9InSZIkST3Md/qkHjSadw77L33HBJZEkiRJ3Walr+bsUESSJEnScGzeKUmSJEk9rDKVvohYFhEPRcSWiFjd7fKofowhjZUxpLEyhjRWxpDGyhhSO5Vo3hkR04DPAm8DtgG3R8SGzHyguyVTXRhDGitjSGNlDGmsjCGNlTG0+3r9lalKVPqAY4AtmfkwQESsB04FDFB1yhjSWBlDGitjaIJ1elFW4w6qjKFBev1CfAIYQ2qrKpW+g4GtLdPbgGMHZ4qIc4BzyuRARDw0CWVr50DgyS797XH1h7u5LvGpMf3ZXxnT0u3VLYaGtLv7ZHeNcV/urrGuY1Vi6OcRcd8ElGXCjCa+uhQb7UzE/0RVYmh3j0NT/jw0nEmK3brH0HiqXDwOFVddPq4NLlPVY6hy+7Vqhjt+TWKsdRRHVan0RZu03CUh80rgyokvzvAi4o7MXNztcoyHHlqXWsXQcHponwypous46hiq6HoMyzJPqEk7DtVom4yol9ZlHNTyXFbFfTiFyzRuMVTFbVg1ddpGVenIZRtwSMv0HOCRLpVF9WQMaayMIY2VMaSxMoY0VsaQ2qpKpe92YF5EHBoRewBnABu6XCbVizGksTKGNFbGkMbKGNJYGUNqqxLNOzNzZ0ScB9wETAPWZOb9XS7WcCrTpGIc9MS61DCGhtMT+2QElVvH3Yyhyq1HByzzBJnk41AttkmHemldxqTG57Iq7sMpWaZxjqEqbsOqqc02isxdmvlKkiRJknpEVZp3SpIkSZImgJU+SZIkSephVvo6FBH/LSK+GxH3RMRXImK/lnkXRcSWiHgoIk7uZjlHIyKWlTJviYjV3S7PVNer+yMi1kTEE63j2UXEARFxc0RsLr/372YZR2OoY0FEzI2IZyPi7vLzV90ua6s6xFdEHBIR34yIByPi/oj4UEn/aET8sGXbntLtsnZDr52H6hCTGlpVj4VViKu6H8uGK2cdjzUTpQqxNhq+09ehiDgJuKW8IPspgMy8MCKOAK4FjgF+GfgH4Ncy88XulXZkETEN+DfgbTS6970dODMzH+hqwaaoXt4fEXECMABck5lHlrRPA9sz89JyoNw/My/sZjk7NcyxYC5wfXMdq6Qu8RURs4HZmXlXRMwA7gTeBbwbGMjMP+lqAbusl85DdYlJDa2Kx8KqxFXdj2UR8VHalLOOx5qJUpVYGw2f9HUoM7+emTvL5CYa454AnAqsz8yfZ+b3gS00/hmq7hhgS2Y+nJnPA+tprIu6o2f3R2b+E7B9UPKpwNryeS2Nk2EtDHMsqLJaxFdmPpqZd5XPO4AHgYO7W6rq6LHzUC1iUkOr6LGwEnHVw8eyOh5rJkolYm00rPTtnvcDXyufDwa2tszbRj3+seta7l411fbHrMx8FBonR2Bml8uzu1qPBQCHRsS3I+IfI+Lfd6tQbdQuvsrTgjcBt5Wk80ozsjV1ag48gep+HqpjmTW0qhwLKxdXNT6WtStn5bZvF9VuW1jpaxER/xAR97X5ObUlz0eAncAXm0ltvqoObWbrWu5e5f6okN08FjwK/LvMfBNwPvA3EbHv5Je+rVrFV0TsA/w98OHM/AnwOeBXgUU0tvNlXSzehJpC56E6lnnKqeGxsFJxVeVj2Qj7dqhyVmr7dlnttkUlBmevisz8zeHmR8RK4J3Aifnyy5DbgENass0BHpmYEo6rupa7V021/fF4RMzOzEfLuw9PdLtArXbnWJCZPwd+Xj7fGRHfA34NuGOCi9uJ2sRXRLyKxkXSFzPzywCZ+XjL/M8D13epeBNuCp2H6ljmKaeGx8LKxFXVj2Uj7dumQeWszPatgNptC5/0dSgilgEXAr+VmT9rmbUBOCMi9oyIQ4F5wLe6UcZRuh2YFxGHRsQewBk01kXdMdX2xwZgZfm8EvhqF8syKkMdCyLioPJiNxHxehrHgoe7U8pd1CK+IiKAq4AHM/NPW9Jnt2T7beC+wctOBT12HqpFTGpoFT0WViKu6n4sG6acdTzWTJRKxNpo+KSvc38B7Anc3PhfZlNm/l5m3h8R1wEP0GjecG4dejEqvW2dB9wETAPWZOb9XS7WlNXL+yMirgWWAAdGxDbgYuBS4LqIWAX8ADi9eyUctbbHAuAE4GMRsRN4Efi9zBzcgU1X1Ci+3gy8F7g3Iu4uaX8EnBkRi2g0nekHPtCd4nVdz5yHahSTGlrljoUViqu6H8s+3a6cdTzWTJQKxVrHHLJBkiRJknqYzTslSZIkqYdZ6ZMkSZKkHmalr0si4qaI+Fib9FMj4rGIODAi1kbEE+Xno10opiRJkqSas9LXPVcD7y09PLV6L42xbv4b8GpgLnBMyfu+ySygJEmSpPqzI5cuiYi9gMeA/5CZ/1TS9qcxCOaxwDeAt2fm7WXeH5Xpf9+lIkuSJEmqIZ/0dUlmPgtcB5zVkvxu4LuZ+Z0y3foUMIAjJ6l4kiRJknqElb7uWgucXp76QaMCuLZ8vhFYHREzIuIw4P00mntKkiRJUses9HVRZv4z8CPg1Ih4PfDrwN+U2X8IPAtsBr4KXAts60Y5JUmSJNXX9G4XQFxD4wnf4cDXM/NxgMzcDqxoZoqIPwa+1ZUSSpIkSaotO3LpsoiYC/wb8ATwf2bm35b0XwWeLj8nAeuAt2Tm/d0pqSRJkqQ6snlnl2VmP/AvwN7AhpZZRwP3AjuAS4AVVvgkSZIkjZZP+iRJkiSph/mkT5IkSZJ6mJU+SZIkSephVvokSZIkqYdZ6ZMkSZKkHlbbcfoOPPDAnDt37m4v/9Of/pS99957/Ao0BlOtLHfeeeeTmXnQhP4RSZIkSUCNK31z587ljjvu2O3l+/r6WLJkyfgVaAymWlki4n9N6B+QJEmS9BKbd0qSJElSD7PSJ0mSJEk9zEqfJEmSJPWw2r7TV0dzV9/QNv2ChTs5u2Ve/6XvmKwiSZIkSepxPumTJEmSpB5mpU+SJEmSepiVPkmSJEnqYVb6JEmSJKmHWemTJEmSpB5mpU+SJEmSepiVPkmSJEnqYVb6JEmSJKmHWemTJEmSpB5mpU+SJEmSepiVPkmSJEnqYR1X+iJiWkR8OyKuL9OHRsRtEbE5Ir4UEXuU9D3L9JYyf27Ld1xU0h+KiJNb0peVtC0RsXr8Vk+SJEmSprbRPOn7EPBgy/SngMszcx7wFLCqpK8CnsrMw4DLSz4i4gjgDGABsAz4y1KRnAZ8Fng7cARwZskrSZIkSRqjjip9ETEHeAfwhTIdwFuBvytZ1gLvKp9PLdOU+SeW/KcC6zPz55n5fWALcEz52ZKZD2fm88D6kleSJEmSNEbTO8z3Z8B/BmaU6dcCT2fmzjK9DTi4fD4Y2AqQmTsj4pmS/2BgU8t3ti6zdVD6se0KERHnAOcAzJo1i76+vg6Lv6uBgYExLb87Lli4s236rL1eOW+yy9WqG9tFkiRJ0sQZsdIXEe8EnsjMOyNiSTO5TdYcYd5Q6e2eNmabNDLzSuBKgMWLF+eSJUvaw46raQAADNJJREFUZetIX18fY1l+d5y9+oa26Rcs3Mll9768K/pXLJmkEu2qG9tFkiRJ0sTp5Enfm4HfiohTgF8C9qXx5G+/iJhenvbNAR4p+bcBhwDbImI68Bpge0t6U+syQ6VLkiRJksZgxHf6MvOizJyTmXNpdMRyS2auAL4JnFayrQS+Wj5vKNOU+bdkZpb0M0rvnocC84BvAbcD80pvoHuUv7FhXNZOkiRJkqa4Tt/pa+dCYH1EfAL4NnBVSb8KWBcRW2g84TsDIDPvj4jrgAeAncC5mfkiQEScB9wETAPWZOb9YyiXJEmSJKkYVaUvM/uAvvL5YRo9bw7O8xxw+hDLfxL4ZJv0jcDG0ZRFkiRJkjSy0YzTJ0mSJEmqGSt9kiRJktTDrPRJkiRJUg+z0idJkiRJPcxKnyRJkiT1MCt9kiRJktTDrPRJkiRJUg+z0idJkiRJPcxKnyRJkiT1MCt9kiRJktTDrPRJkiRJUg+z0idJkiRJPcxKnyRJkiT1sBErfRHxSxHxrYj4TkTcHxH/T0k/NCJui4jNEfGliNijpO9ZpreU+XNbvuuikv5QRJzckr6spG2JiNXjv5qSJEmSNDV18qTv58BbM/ONwCJgWUQcB3wKuDwz5wFPAatK/lXAU5l5GHB5yUdEHAGcASwAlgF/GRHTImIa8Fng7f9/e3cXYtd1nnH8/9SOi6Cklpt6KiTRGCpK1AiaZJBVejPUiS0nF3JLDTamloNhIMi0Bd2o7YUhaUC9cC8MaUBgYbmUCNEPLGqnQlUzmIDdyP3AsiyMpo6IBgmbosT1EJow6duLs4SPR2c0oznzuef/g8M5+91r7f2eM1cPe+81wE7g0TZWkiRJkjSkeUNf9Uy3zY+1VwG/A/xtqx8DHmqf97Vt2v77kqTVj1fVT6rq+8AksLu9Jqvqnar6KXC8jZUkSZIkDen2hQxqV+P+Dfg1elfl/gv4UVXNtCFTwNb2eStwGaCqZpK8D/xSq7/Wd9j+OZdn1e+do49xYBxgZGSEiYmJhbQ/0PT09FDzF+PgrpmB9ZFNH9230n31W43fRZIkSdLyWVDoq6qfAb+Z5E7gH4BPDRrW3jPHvrnqg6421oAaVXUEOAIwOjpaY2NjN2/8JiYmJhhm/mI8ceilgfWDu2Z45tyHf4pLj42tUEc3Wo3fRZIkSdLyuaXVO6vqR8AEsAe4M8n1pLINuNI+TwHbAdr+XwSu9ddnzZmrLkmSJEka0kJW7/zldoWPJJuAzwMXgO8Av9+G7QdebJ9Ptm3a/n+pqmr1R9rqnvcAO4DvAWeBHW010DvoLfZycim+nCRJkiRtdAu5vXMLcKw91/dzwImq+sckbwHHk/w58B/Ac238c8BfJ5mkd4XvEYCqOp/kBPAWMAMcaLeNkuQp4BRwG3C0qs4v2TeUJEmSpA1s3tBXVW8AnxlQf4feypuz6/8LPDzHsb4OfH1A/WXg5QX0K0mSJEm6Bbf0TJ8kSZIkaX0x9EmSJElShxn6JEmSJKnDDH2SJEmS1GGGPkmSJEnqMEOfJEmSJHWYoU+SJEmSOszQJ0mSJEkdZuiTJEmSpA4z9EmSJElShxn6JEmSJKnDDH2SJEmS1GGGPkmSJEnqsHlDX5LtSb6T5EKS80n+qNXvSnI6ycX2vrnVk+TZJJNJ3kjy2b5j7W/jLybZ31f/XJJzbc6zSbIcX1aSJEmSNpqFXOmbAQ5W1aeAPcCBJDuBQ8CZqtoBnGnbAA8CO9prHPgm9EIi8DRwL7AbePp6UGxjxvvm7R3+q0mSJEmS5g19VXW1qv69ff4AuABsBfYBx9qwY8BD7fM+4IXqeQ24M8kW4AHgdFVdq6ofAqeBvW3fx6vq1aoq4IW+Y0mSJEmShnD7rQxO8kngM8C/AiNVdRV6wTDJ3W3YVuBy37SpVrtZfWpAfdD5x+ldEWRkZISJiYlbaf8jpqenh5q/GAd3zQysj2z66L6V7qvfavwukiRJkpbPgkNfkl8A/g7446r6n5s8djdoRy2ifmOx6ghwBGB0dLTGxsbm6XpuExMTDDN/MZ449NLA+sFdMzxz7sM/xaXHxlaooxutxu8iSZIkafksaPXOJB+jF/j+pqr+vpXfbbdm0t7fa/UpYHvf9G3AlXnq2wbUJUmSJElDWsjqnQGeAy5U1V/27ToJXF+Bcz/wYl/98baK5x7g/XYb6Cng/iSb2wIu9wOn2r4Pkuxp53q871iSJEmSpCEs5PbO3wb+ADiX5D9b7U+Bw8CJJE8CPwAebvteBr4ITAI/Br4MUFXXknwNONvGfbWqrrXPXwGeBzYB324vSZIkSdKQ5g19VfVdBj93B3DfgPEFHJjjWEeBowPqrwOfnq8XSZIkSdKtWdAzfZIkSZKk9cnQJ0mSJEkdZuiTJEmSpA4z9EmSJElShxn6JEmSJKnDDH2SJEmS1GGGPkmSJEnqMEOfJEmSJHWYoU+SJEmSOszQJ0mSJEkdZuiTJEmSpA4z9EmSJElShxn6JEmSJKnD5g19SY4meS/Jm321u5KcTnKxvW9u9SR5NslkkjeSfLZvzv42/mKS/X31zyU51+Y8myRL/SUlSZIkaaNayJW+54G9s2qHgDNVtQM407YBHgR2tNc48E3ohUTgaeBeYDfw9PWg2MaM982bfS5JkiRJ0iLNG/qq6hXg2qzyPuBY+3wMeKiv/kL1vAbcmWQL8ABwuqquVdUPgdPA3rbv41X1alUV8ELfsSRJkiRJQ1rsM30jVXUVoL3f3epbgct946Za7Wb1qQF1SZIkSdISuH2JjzfoebxaRH3wwZNxereCMjIywsTExCJa7Jmenh5q/mIc3DUzsD6y6aP7Vrqvfqvxu0iSJElaPosNfe8m2VJVV9stmu+1+hSwvW/cNuBKq4/Nqk+0+rYB4weqqiPAEYDR0dEaGxuba+i8JiYmGGb+Yjxx6KWB9YO7Znjm3Id/ikuPja1QRzdajd9FkiRJ0vJZ7O2dJ4HrK3DuB17sqz/eVvHcA7zfbv88BdyfZHNbwOV+4FTb90GSPW3Vzsf7jiVJkiRJGtK8V/qSfIveVbpPJJmitwrnYeBEkieBHwAPt+EvA18EJoEfA18GqKprSb4GnG3jvlpV1xeH+Qq9FUI3Ad9uL0mSJEnSEpg39FXVo3Psum/A2AIOzHGco8DRAfXXgU/P14ckSZIk6dYt9vZOSZIkSdI6sNSrd2oJfHKOBV8GuXT4S8vYiSRJkqT1zit9kiRJktRhhj5JkiRJ6jBDnyRJkiR1mKFPkiRJkjrM0CdJkiRJHWbokyRJkqQOM/RJkiRJUocZ+iRJkiSpw/zn7EvgVv6ZuiRJkiStJK/0SZIkSVKHGfokSZIkqcMMfZIkSZLUYWsm9CXZm+TtJJNJDq12P5IkSZLUBWsi9CW5DfgG8CCwE3g0yc7V7UqSJEmS1r+1snrnbmCyqt4BSHIc2Ae8tapdrQMLXTn00uEvLXMnkiRJktaitRL6tgKX+7angHtnD0oyDoy3zekkbw9xzk8A/z3E/CXzhyvQS/5iwUNX4nf51WU+viRJkqRmrYS+DKjVDYWqI8CRJTlh8npVjS7FsYZlL5IkSZKWy5p4po/elb3tfdvbgCur1IskSZIkdcZaCX1ngR1J7klyB/AIcHKVe5IkSZKkdW9N3N5ZVTNJngJOAbcBR6vq/DKfdkluE10i9iJJkiRpWaTqhkfnJEmSJEkdsVZu75QkSZIkLQNDnyRJkiR12IYLfUkeTnI+yf8lGZ2170+STCZ5O8kDK9TP3na+ySSHVuKcfec+muS9JG/21e5KcjrJxfa+eSV7kiRJkrS0NlzoA94Efg94pb+YZCe9VUN/A9gL/FWS25azkXb8bwAPAjuBR1sfK+V5et+13yHgTFXtAM60bUmSJEnr1IYLfVV1oareHrBrH3C8qn5SVd8HJoHdy9zObmCyqt6pqp8Cx1sfK6KqXgGuzSrvA461z8eAh1aqH0mSJElLb8OFvpvYClzu255qta6dcz4jVXUVoL3fvcr9SJIkSRrCmvg/fUstyT8DvzJg159V1YtzTRtQW+7/Z7Ea55QkSZK0gXQy9FXV5xcxbQrY3re9DbiyNB2tqXPO590kW6rqapItwHur3I8kSZKkIXh754dOAo8k+fkk9wA7gO8t8znPAjuS3JPkDnoLyZxc5nPO5ySwv33eD8x1ZVSSJEnSOrDhQl+S300yBfwW8FKSUwBVdR44AbwF/BNwoKp+tpy9VNUM8BRwCrgAnGh9rIgk3wJeBX49yVSSJ4HDwBeSXAS+0LYlSZIkrVOp8hEySZIkSeqqDXelT5IkSZI2EkOfJEmSJHWYoU+SJEmSOszQJ0mSJEkdZuiTJEmSpA4z9EmSJElShxn6JEmSJKnD/h/knEolau5PfQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accb4b83d68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.hist(figsize=(15,15))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. Check the spread of fraud vs non-fraud on selected variables\n",
    "#### This is done to identify variables selection which has good spread of frauds and non frauds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "V9 - V10\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accb4b51ef0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accbbbe2dd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sample = df[df['Class']==0]\n",
    "fraud = df[df['Class']==1]\n",
    "print('V9 - V10')\n",
    "plt.scatter(fraud['V9'], fraud['V10'],s=1, color='r')\n",
    "plt.scatter(sample['V9'], sample['V10'], s=1, color='g')\n",
    "plt.show()\n",
    "plt.clf()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "V16-V17\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accbb8fe438>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accbbf81f28>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('V16-V17')\n",
    "plt.scatter(sample['V16'], sample['V17'], s=1, color = 'g')\n",
    "plt.scatter(fraud['V16'], fraud['V17'], s=1, color = 'r')\n",
    "plt.show()\n",
    "plt.clf()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "V17 - V18\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accbbca5208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accbb9762e8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('V17 - V18')\n",
    "plt.scatter(sample['V18'], sample['V17'], s=1, color = 'g')\n",
    "plt.scatter(fraud['V18'], fraud['V17'], s=1, color = 'r')\n",
    "plt.show()\n",
    "plt.clf()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "V1 - V3\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accbbf79eb8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accbb814400>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('V1 - V3')\n",
    "plt.scatter(sample['V1'], sample['V3'], s=1, color = 'g')\n",
    "plt.scatter(fraud['V1'], fraud['V3'], s=1, color = 'r')\n",
    "plt.show()\n",
    "plt.clf()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "V1 - V2\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accbbbe7a58>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2accbbfa6c50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('V1 - V2')\n",
    "plt.scatter(sample['V1'], sample['V2'], s=1, color = 'g')\n",
    "plt.scatter(fraud['V1'], fraud['V2'], s=1, color = 'r')\n",
    "plt.show()\n",
    "plt.clf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 7. Split the data into train & test data sets using 70:30 mix\n",
    "#### The model will be built on training data and will be applied on the test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "xtrain shape\n",
      "(39811, 30)\n",
      "xtest shape\n",
      "(17063, 30)\n"
     ]
    }
   ],
   "source": [
    "# Split The Data with all variables\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "x = df[['Time', 'V1', 'V2', 'V3', 'V4', 'V5', 'V6', 'V7', 'V8', 'V9', 'V10',\n",
    "       'V11', 'V12', 'V13', 'V14', 'V15', 'V16', 'V17', 'V18', 'V19', 'V20',\n",
    "       'V21', 'V22', 'V23', 'V24', 'V25', 'V26', 'V27', 'V28', 'Amount']]\n",
    "y = df['Class']\n",
    "\n",
    "xtrain, xtest, ytrain, ytest = train_test_split(x, y, test_size=0.30, random_state=0)\n",
    "print('xtrain shape')\n",
    "print(xtrain.shape)\n",
    "print('xtest shape')\n",
    "print(xtest.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 8. Use Random Forest Algorithm\n",
    "#### A brief about Random Forest Algorithm\n",
    "Random forest classifier creates a set of decision trees from randomly selected subset of training set. It then aggregates the votes from different decision trees to decide the final class of the test object.  Random forests or random decision forests are an ensemble learning method for classification, regression and other tasks, that operate by constructing a multitude of decision trees at training time and outputting the class that is the mode of the classes (classification) or mean prediction (regression) of the individual trees. Random decision forests will avoid overfitting which will enhance the accuracy of the model on new data. This is a Bagging based algorithm which is used for reducing Overfitting in order to create strong learners for generating accurate predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model\n",
      "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
      "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=1, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1,\n",
      "            oob_score=False, random_state=None, verbose=0,\n",
      "            warm_start=False)\n",
      "confusion matrix\n",
      "[[17029     1]\n",
      " [    9    24]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00     17030\n",
      "          1       0.96      0.73      0.83        33\n",
      "\n",
      "avg / total       1.00      1.00      1.00     17063\n",
      "\n",
      "Accuracy : 0.999414\n",
      "Area under the curve : 0.863607\n"
     ]
    }
   ],
   "source": [
    "# RF Classifier explained\n",
    "\n",
    "'''rf = RandomForestClassifier(n_estimators=100, oob_score=True, n_jobs=4)\n",
    "n_estimators : integer, optional (default=10)\n",
    "The number of trees in the forest.\n",
    "oob_score : bool (default=False)\n",
    "Whether to use out-of-bag samples to estimate the generalization accuracy.\n",
    "n_jobs : integer, optional (default=1)\n",
    "The number of jobs to run in parallel for both fit and predict. If -1, then the number of jobs is set to the number of cores.\n",
    "'''\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn import metrics\n",
    "\n",
    "rfmodel = RandomForestClassifier()\n",
    "rfmodel.fit(xtrain,ytrain)\n",
    "print('model')\n",
    "print(rfmodel)\n",
    "\n",
    "ypredrf = rfmodel.predict(xtest)\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, ypredrf))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, ypredrf))\n",
    "print('Accuracy : %f' % (metrics.accuracy_score(ytest, ypredrf)))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, ypredrf)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### In the above classification report we are getting the f1-score of 89 which is the harmonic mean of precision & recall scores.\n",
    "##### Recall is where the model tries to recollect the number of instances, in this case the model has been able to recollect 82% of frauds and able to classify them as frauds 96 % of the recollected events which is the precision of the model. \n",
    "##### Area under the curve signifies the accuracy of the model (values between 0 & 1) where the score towards 1 indicate high predictive power.\n",
    "# <font color='Red'> Note  : The numbers for recall, precision, F1 Score & Area under the curve can change for different runs of the model due to stocastic nature of the algorithms."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 9. Use Gradient Boosting Alogrithm\n",
    "#### A Brief about Gradient Boosting Algorithm\n",
    "Gradient boosting is a machine learning technique for regression and classification problems, which produces a prediction model in the form of an ensemble of weak prediction models, typically decision trees. It builds the model in a stage-wise fashion like other boosting methods do, and it generalizes them by allowing optimization of an arbitrary differentiable loss function(a loss function or cost function is a function that maps an event or values of one or more variables onto a real number intuitively representing some \"cost\" associated with the event). This is a Boosting based algorithm which is an ensemble technique to combine weak learners to create a strong learner that can make accurate predictions. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "confusion matrix\n",
      "[[17022     8]\n",
      " [   10    23]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00     17030\n",
      "          1       0.74      0.70      0.72        33\n",
      "\n",
      "avg / total       1.00      1.00      1.00     17063\n",
      "\n",
      "-----------------------------------------------------------------------------------------\n",
      "Accuracy is :\n",
      "0.998945085858\n",
      "Area under the curve : 0.848250\n"
     ]
    }
   ],
   "source": [
    "# GBM Classifier explained\n",
    "\n",
    "'''params = {'n_estimators': 500, 'max_depth': 3, 'subsample': 0.5, 'learning_rate': 0.01, 'min_samples_leaf': 1, 'random_state': 3}\n",
    "clf = ensemble. GradientBoostingClassifier(**params)\n",
    "n_estimators : int (default=100)\n",
    "The number of boosting stages to perform. Gradient boosting is fairly robust to over-fitting, so a large number usually results in better performance.\n",
    "max_depth: integer, optional (default=3)\n",
    "maximum depth of the individual regression estimators. The maximum depth limits the number of nodes in the tree. The best value depends on the interaction of the input variables.\n",
    "subsample: float, optional (default=1.0)\n",
    "The fraction of samples to be used for fitting the individual base learners. If smaller than 1.0 this results in Stochastic Gradient Boosting. subsample interacts with the parameter n_estimators. Choosing subsample < 1.0 leads to a reduction of variance and an increase in bias.\n",
    "random_state : int, RandomState instance or None, optional (default=None)\n",
    "If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by np.random.\n",
    "learning_rate : float, optional (default=0.1)\n",
    "learning rate shrinks the contribution of each tree by learning_rate. There is a trade-off between learning_rate and n_estimators.\n",
    "min_samples_leaf : int, float, optional (default=1)\n",
    "The minimum number of samples required to be at a leaf node:\n",
    "'''\n",
    "\n",
    "from sklearn import ensemble\n",
    "\n",
    "params = {'n_estimators': 500, 'max_depth': 3, 'subsample': 0.5,\n",
    "          'learning_rate': 0.01, 'min_samples_leaf': 1, 'random_state': 3}\n",
    "clf = ensemble.GradientBoostingClassifier(**params)\n",
    "clf.fit(xtrain, ytrain) #trains\n",
    "y_pred = clf.predict(xtest)  #predicts\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, y_pred))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, y_pred))\n",
    "print(\"-----------------------------------------------------------------------------------------\")\n",
    "print(\"Accuracy is :\")\n",
    "print(metrics.accuracy_score(ytest, y_pred))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, y_pred)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 10. Use Extreme Gradient Boosting Alogrithm\n",
    "#### A Brief about Extreme Gradient Boosting Algorithm\n",
    "XGBoost is one of the implementations of Gradient Boosting concept, but what makes XGBoost unique is that it uses “a more regularized model formalization to control over-fitting, which gives it better performance,” according to the author of the algorithm, Tianqi Chen. Therefore, it helps to reduce overfitting."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/DSX-Python35/lib/python3.5/site-packages/sklearn/cross_validation.py:41: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.\n",
      "  \"This module will be removed in 0.20.\", DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "from xgboost.sklearn import XGBClassifier\n",
    "\n",
    "# Create the XGB classifier, xgb_model.\n",
    "xgb_model = XGBClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'reg_alpha': 0, 'reg_lambda': 1, 'max_depth': 3, 'scale_pos_weight': 1, 'n_estimators': 100, 'max_delta_step': 0, 'min_child_weight': 1, 'seed': 0, 'colsample_bylevel': 1, 'colsample_bytree': 1, 'silent': 1, 'subsample': 1, 'learning_rate': 0.1, 'objective': 'binary:logistic', 'base_score': 0.5, 'gamma': 0, 'missing': None}\n"
     ]
    }
   ],
   "source": [
    "# List the default parameters.\n",
    "print(xgb_model.get_xgb_params())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\tvalidation_0-error:0.000502\tvalidation_1-error:0.000762\n",
      "[1]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000762\n",
      "[2]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000762\n",
      "[3]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000762\n",
      "[4]\tvalidation_0-error:0.000502\tvalidation_1-error:0.000645\n",
      "[5]\tvalidation_0-error:0.000427\tvalidation_1-error:0.00082\n",
      "[6]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000762\n",
      "[7]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000762\n",
      "[8]\tvalidation_0-error:0.000352\tvalidation_1-error:0.000703\n",
      "[9]\tvalidation_0-error:0.000352\tvalidation_1-error:0.000762\n",
      "[10]\tvalidation_0-error:0.000352\tvalidation_1-error:0.000703\n",
      "[11]\tvalidation_0-error:0.000327\tvalidation_1-error:0.000527\n",
      "[12]\tvalidation_0-error:0.000327\tvalidation_1-error:0.000527\n",
      "[13]\tvalidation_0-error:0.000301\tvalidation_1-error:0.000586\n",
      "[14]\tvalidation_0-error:0.000301\tvalidation_1-error:0.000527\n",
      "[15]\tvalidation_0-error:0.000301\tvalidation_1-error:0.000586\n",
      "[16]\tvalidation_0-error:0.000301\tvalidation_1-error:0.000527\n",
      "[17]\tvalidation_0-error:0.000301\tvalidation_1-error:0.000586\n",
      "[18]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000469\n",
      "[19]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000527\n",
      "[20]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000527\n",
      "[21]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000586\n",
      "[22]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000527\n",
      "[23]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000586\n",
      "[24]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000586\n",
      "[25]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000527\n",
      "[26]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000527\n",
      "[27]\tvalidation_0-error:0.000251\tvalidation_1-error:0.000527\n",
      "[28]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000527\n",
      "[29]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000586\n",
      "[30]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000527\n",
      "[31]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000586\n",
      "[32]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000586\n",
      "[33]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000586\n",
      "[34]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000527\n",
      "[35]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000586\n",
      "[36]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000586\n",
      "[37]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000527\n",
      "[38]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000527\n",
      "[39]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000586\n",
      "[40]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000469\n",
      "[41]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000469\n",
      "[42]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000469\n",
      "[43]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000469\n",
      "[44]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000469\n",
      "[45]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000469\n",
      "[46]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000469\n",
      "[47]\tvalidation_0-error:0.000226\tvalidation_1-error:0.000527\n",
      "[48]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[49]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[50]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[51]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[52]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[53]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[54]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[55]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[56]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[57]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[58]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[59]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[60]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[61]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[62]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[63]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[64]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[65]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[66]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[67]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[68]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[69]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[70]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[71]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[72]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[73]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[74]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[75]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[76]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[77]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[78]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[79]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[80]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[81]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[82]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[83]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[84]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[85]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[86]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[87]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[88]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[89]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[90]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[91]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[92]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[93]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[94]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[95]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[96]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[97]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[98]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n",
      "[99]\tvalidation_0-error:0.000201\tvalidation_1-error:0.000527\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "XGBClassifier(base_score=0.5, colsample_bylevel=1, colsample_bytree=1,\n",
       "       gamma=0, learning_rate=0.1, max_delta_step=0, max_depth=3,\n",
       "       min_child_weight=1, missing=None, n_estimators=100, nthread=-1,\n",
       "       objective='binary:logistic', reg_alpha=0, reg_lambda=1,\n",
       "       scale_pos_weight=1, seed=0, silent=True, subsample=1)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train and evaluate.\n",
    "xgb_model.fit(xtrain, ytrain, eval_metric=['error'], eval_set=[((xtrain, ytrain)),(xtest, ytest)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "confusion matrix\n",
      "[[17028     2]\n",
      " [    7    26]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00     17030\n",
      "          1       0.93      0.79      0.85        33\n",
      "\n",
      "avg / total       1.00      1.00      1.00     17063\n",
      "\n",
      "-----------------------------------------------------------------------------------------\n",
      "Accuracy is :\n",
      "0.999472542929\n",
      "Area under the curve : 0.893881\n"
     ]
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "y_pred = xgb_model.predict(xtest)  #predicts\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, y_pred))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, y_pred))\n",
    "print(\"-----------------------------------------------------------------------------------------\")\n",
    "print(\"Accuracy is :\")\n",
    "print(metrics.accuracy_score(ytest, y_pred))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, y_pred)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### As we can see, the F1 score and Area under the curve are higher compared to previous model. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 11. Select random variables for model building\n",
    "#### This is done to check whether we can get a lift on the accuracy with fewer variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "xtrain shape\n",
      "(39811, 6)\n",
      "xtest shape\n",
      "(17063, 6)\n"
     ]
    }
   ],
   "source": [
    "# Split The Data with few variables\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "x = df[['V9', 'V10','V16', 'V17', 'V18', 'Amount']]\n",
    "y = df['Class']\n",
    "\n",
    "xtrain, xtest, ytrain, ytest = train_test_split(x, y, test_size=0.30, random_state=0)\n",
    "print('xtrain shape')\n",
    "print(xtrain.shape)\n",
    "print('xtest shape')\n",
    "print(xtest.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model\n",
      "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
      "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=1, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1,\n",
      "            oob_score=False, random_state=None, verbose=0,\n",
      "            warm_start=False)\n",
      "confusion matrix\n",
      "[[17028     2]\n",
      " [    8    25]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00     17030\n",
      "          1       0.93      0.76      0.83        33\n",
      "\n",
      "avg / total       1.00      1.00      1.00     17063\n",
      "\n",
      "Accuracy : 0.999414\n",
      "Area under the curve : 0.878729\n"
     ]
    }
   ],
   "source": [
    "'''Random Forest Classifier on reduced dimensions data'''\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn import metrics\n",
    "\n",
    "rfmodel = RandomForestClassifier()\n",
    "rfmodel.fit(xtrain,ytrain)\n",
    "print('model')\n",
    "print(rfmodel)\n",
    "\n",
    "ypredrf = rfmodel.predict(xtest)\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, ypredrf))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, ypredrf))\n",
    "print('Accuracy : %f' % (metrics.accuracy_score(ytest, ypredrf)))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, ypredrf)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### The F1 score has reduced with fewer variables along with Area under the curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "confusion matrix\n",
      "[[17023     7]\n",
      " [    8    25]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00     17030\n",
      "          1       0.78      0.76      0.77        33\n",
      "\n",
      "avg / total       1.00      1.00      1.00     17063\n",
      "\n",
      "-----------------------------------------------------------------------------------------\n",
      "Accuracy is :\n",
      "0.999120904882\n",
      "Area under the curve : 0.878582\n"
     ]
    }
   ],
   "source": [
    "'''Gradient Boost Algorithm on reduced dimensions data'''\n",
    "\n",
    "from sklearn import ensemble\n",
    "\n",
    "params = {'n_estimators': 500, 'max_depth': 3, 'subsample': 0.5,\n",
    "          'learning_rate': 0.01, 'min_samples_leaf': 1, 'random_state': 3}\n",
    "clf = ensemble.GradientBoostingClassifier(**params)\n",
    "clf.fit(xtrain, ytrain) #trains\n",
    "y_pred = clf.predict(xtest)  #predicts\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, y_pred))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, y_pred))\n",
    "print(\"-----------------------------------------------------------------------------------------\")\n",
    "print(\"Accuracy is :\")\n",
    "print(metrics.accuracy_score(ytest, y_pred))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, y_pred)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Reduction in the F1 score with fewer variables."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'reg_alpha': 0, 'reg_lambda': 1, 'max_depth': 3, 'scale_pos_weight': 1, 'n_estimators': 100, 'max_delta_step': 0, 'min_child_weight': 1, 'seed': 0, 'colsample_bylevel': 1, 'colsample_bytree': 1, 'silent': 1, 'subsample': 1, 'learning_rate': 0.1, 'objective': 'binary:logistic', 'base_score': 0.5, 'gamma': 0, 'missing': None}\n"
     ]
    }
   ],
   "source": [
    "'''Extreme Gradient Boost Algorithm on reduced dimensions data'''\n",
    "\n",
    "from xgboost.sklearn import XGBClassifier\n",
    "\n",
    "# Create the XGB classifier, xgb_model.\n",
    "xgb_model = XGBClassifier()\n",
    "# List the default parameters.\n",
    "print(xgb_model.get_xgb_params())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000645\n",
      "[1]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[2]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[3]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[4]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[5]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[6]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[7]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[8]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[9]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[10]\tvalidation_0-error:0.000553\tvalidation_1-error:0.000586\n",
      "[11]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000586\n",
      "[12]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000645\n",
      "[13]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000586\n",
      "[14]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000645\n",
      "[15]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000645\n",
      "[16]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000586\n",
      "[17]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000645\n",
      "[18]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000645\n",
      "[19]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000645\n",
      "[20]\tvalidation_0-error:0.000527\tvalidation_1-error:0.000645\n",
      "[21]\tvalidation_0-error:0.000477\tvalidation_1-error:0.000527\n",
      "[22]\tvalidation_0-error:0.000477\tvalidation_1-error:0.000527\n",
      "[23]\tvalidation_0-error:0.000477\tvalidation_1-error:0.000527\n",
      "[24]\tvalidation_0-error:0.000477\tvalidation_1-error:0.000527\n",
      "[25]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[26]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[27]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[28]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[29]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[30]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[31]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[32]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[33]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[34]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[35]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[36]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[37]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[38]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[39]\tvalidation_0-error:0.000427\tvalidation_1-error:0.000527\n",
      "[40]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[41]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[42]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[43]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[44]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[45]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[46]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[47]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[48]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[49]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[50]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[51]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[52]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[53]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[54]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[55]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[56]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[57]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[58]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[59]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[60]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[61]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[62]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[63]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[64]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[65]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[66]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[67]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[68]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[69]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[70]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[71]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[72]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[73]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[74]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[75]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[76]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[77]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[78]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[79]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[80]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[81]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[82]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000527\n",
      "[83]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[84]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[85]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[86]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[87]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[88]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[89]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[90]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[91]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[92]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[93]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[94]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[95]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[96]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[97]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[98]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n",
      "[99]\tvalidation_0-error:0.000402\tvalidation_1-error:0.000586\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "XGBClassifier(base_score=0.5, colsample_bylevel=1, colsample_bytree=1,\n",
       "       gamma=0, learning_rate=0.1, max_delta_step=0, max_depth=3,\n",
       "       min_child_weight=1, missing=None, n_estimators=100, nthread=-1,\n",
       "       objective='binary:logistic', reg_alpha=0, reg_lambda=1,\n",
       "       scale_pos_weight=1, seed=0, silent=True, subsample=1)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train and evaluate.\n",
    "xgb_model.fit(xtrain, ytrain, eval_metric=['error'], eval_set=[((xtrain, ytrain)),(xtest, ytest)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "confusion matrix\n",
      "[[17027     3]\n",
      " [    7    26]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00     17030\n",
      "          1       0.90      0.79      0.84        33\n",
      "\n",
      "avg / total       1.00      1.00      1.00     17063\n",
      "\n",
      "-----------------------------------------------------------------------------------------\n",
      "Accuracy is :\n",
      "0.999413936588\n",
      "Area under the curve : 0.893851\n"
     ]
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "y_pred = xgb_model.predict(xtest)  #predicts\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, y_pred))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, y_pred))\n",
    "print(\"-----------------------------------------------------------------------------------------\")\n",
    "print(\"Accuracy is :\")\n",
    "print(metrics.accuracy_score(ytest, y_pred))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, y_pred)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Higher F1 score compared to Gradient Boosting Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 11. Random Under Sampling\n",
    "#### This is done to reduce the imbalance between frauds & non frauds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "'''We are taking 10% of non frauds and merge it with frauds data'''\n",
    "\n",
    "df = df[['Class', 'Time', 'V1', 'V2', 'V3', 'V4', 'V5', 'V6', 'V7', 'V8', 'V9', 'V10',\n",
    "       'V11', 'V12', 'V13', 'V14', 'V15', 'V16', 'V17', 'V18', 'V19', 'V20',\n",
    "       'V21', 'V22', 'V23', 'V24', 'V25', 'V26', 'V27', 'V28', 'Amount']]\n",
    "\n",
    "\n",
    "sample = df[df['Class']==0]\n",
    "fraud = df[df['Class'] == 1]\n",
    "\n",
    "# random sampling\n",
    "ignore_me, sample = train_test_split(sample, test_size = 0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "sample = pd.concat([sample, fraud])\n",
    "\n",
    "# Split into train and test units.\n",
    "xtrain, xtest = train_test_split(sample, test_size = 0.3)\n",
    "ytrain = xtrain['Class']\n",
    "ytest = xtest['Class']\n",
    "xtrain.drop('Class', 1, inplace = True)\n",
    "xtest.drop('Class', 1, inplace = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model\n",
      "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
      "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=1, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1,\n",
      "            oob_score=False, random_state=None, verbose=0,\n",
      "            warm_start=False)\n",
      "confusion matrix\n",
      "[[1703    0]\n",
      " [   2   29]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00      1703\n",
      "          1       1.00      0.94      0.97        31\n",
      "\n",
      "avg / total       1.00      1.00      1.00      1734\n",
      "\n",
      "Accuracy : 0.998847\n",
      "Area under the curve : 0.967742\n"
     ]
    }
   ],
   "source": [
    "'''Random Forest Classifier on undersampled data'''\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn import metrics\n",
    "\n",
    "rfmodel = RandomForestClassifier()\n",
    "rfmodel.fit(xtrain,ytrain)\n",
    "print('model')\n",
    "print(rfmodel)\n",
    "\n",
    "ypredrf = rfmodel.predict(xtest)\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, ypredrf))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, ypredrf))\n",
    "print('Accuracy : %f' % (metrics.accuracy_score(ytest, ypredrf)))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, ypredrf)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### we can see an increase in F1 score which indicates that random under sample can give a lift in the accuracy. However we need to do random sampling multiple times for training the model and then use it on the new data for better results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "confusion matrix\n",
      "[[1700    3]\n",
      " [   3   28]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00      1703\n",
      "          1       0.90      0.90      0.90        31\n",
      "\n",
      "avg / total       1.00      1.00      1.00      1734\n",
      "\n",
      "-----------------------------------------------------------------------------------------\n",
      "Accuracy is :\n",
      "0.996539792388\n",
      "Area under the curve : 0.950732\n"
     ]
    }
   ],
   "source": [
    "'''Gradient Boost Algorithm on undersampled data'''\n",
    "\n",
    "from sklearn import ensemble\n",
    "\n",
    "params = {'n_estimators': 650, 'max_depth': 3, 'subsample': 0.5,\n",
    "          'learning_rate': 0.01, 'min_samples_leaf': 1, 'random_state': 3}\n",
    "clf = ensemble.GradientBoostingClassifier(**params)\n",
    "clf.fit(xtrain, ytrain) #trains\n",
    "y_pred = clf.predict(xtest)  #predicts\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, y_pred))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, y_pred))\n",
    "print(\"-----------------------------------------------------------------------------------------\")\n",
    "print(\"Accuracy is :\")\n",
    "print(metrics.accuracy_score(ytest, y_pred))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, y_pred)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### There is an increase in the F1 score for this sampling technique with 10% data of non-frauds and 100% frauds. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'reg_alpha': 0, 'reg_lambda': 1, 'max_depth': 3, 'scale_pos_weight': 1, 'n_estimators': 100, 'max_delta_step': 0, 'min_child_weight': 1, 'seed': 0, 'colsample_bylevel': 1, 'colsample_bytree': 1, 'silent': 1, 'subsample': 1, 'learning_rate': 0.1, 'objective': 'binary:logistic', 'base_score': 0.5, 'gamma': 0, 'missing': None}\n"
     ]
    }
   ],
   "source": [
    "'''Extreme Gradient Boost Algorithm on reduced dimensions data'''\n",
    "\n",
    "from xgboost.sklearn import XGBClassifier\n",
    "\n",
    "# Create the XGB classifier, xgb_model.\n",
    "xgb_model = XGBClassifier()\n",
    "\n",
    "# List the default parameters.\n",
    "print(xgb_model.get_xgb_params())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[1]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[2]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[3]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[4]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[5]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[6]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[7]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[8]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[9]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[10]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[11]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[12]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[13]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[14]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[15]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[16]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[17]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[18]\tvalidation_0-error:0.002472\tvalidation_1-error:0.002307\n",
      "[19]\tvalidation_0-error:0.002472\tvalidation_1-error:0.00173\n",
      "[20]\tvalidation_0-error:0.002472\tvalidation_1-error:0.00173\n",
      "[21]\tvalidation_0-error:0.002472\tvalidation_1-error:0.00173\n",
      "[22]\tvalidation_0-error:0.002472\tvalidation_1-error:0.00173\n",
      "[23]\tvalidation_0-error:0.002472\tvalidation_1-error:0.00173\n",
      "[24]\tvalidation_0-error:0.002472\tvalidation_1-error:0.00173\n",
      "[25]\tvalidation_0-error:0.002224\tvalidation_1-error:0.001153\n",
      "[26]\tvalidation_0-error:0.002224\tvalidation_1-error:0.001153\n",
      "[27]\tvalidation_0-error:0.002224\tvalidation_1-error:0.001153\n",
      "[28]\tvalidation_0-error:0.002224\tvalidation_1-error:0.001153\n",
      "[29]\tvalidation_0-error:0.002224\tvalidation_1-error:0.001153\n",
      "[30]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[31]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[32]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[33]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[34]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[35]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[36]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[37]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[38]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[39]\tvalidation_0-error:0.001977\tvalidation_1-error:0.001153\n",
      "[40]\tvalidation_0-error:0.00173\tvalidation_1-error:0.001153\n",
      "[41]\tvalidation_0-error:0.00173\tvalidation_1-error:0.001153\n",
      "[42]\tvalidation_0-error:0.00173\tvalidation_1-error:0.001153\n",
      "[43]\tvalidation_0-error:0.00173\tvalidation_1-error:0.001153\n",
      "[44]\tvalidation_0-error:0.00173\tvalidation_1-error:0.001153\n",
      "[45]\tvalidation_0-error:0.00173\tvalidation_1-error:0.001153\n",
      "[46]\tvalidation_0-error:0.00173\tvalidation_1-error:0.00173\n",
      "[47]\tvalidation_0-error:0.00173\tvalidation_1-error:0.00173\n",
      "[48]\tvalidation_0-error:0.00173\tvalidation_1-error:0.00173\n",
      "[49]\tvalidation_0-error:0.00173\tvalidation_1-error:0.001153\n",
      "[50]\tvalidation_0-error:0.00173\tvalidation_1-error:0.00173\n",
      "[51]\tvalidation_0-error:0.00173\tvalidation_1-error:0.00173\n",
      "[52]\tvalidation_0-error:0.00173\tvalidation_1-error:0.00173\n",
      "[53]\tvalidation_0-error:0.00173\tvalidation_1-error:0.001153\n",
      "[54]\tvalidation_0-error:0.001483\tvalidation_1-error:0.00173\n",
      "[55]\tvalidation_0-error:0.001483\tvalidation_1-error:0.001153\n",
      "[56]\tvalidation_0-error:0.001483\tvalidation_1-error:0.001153\n",
      "[57]\tvalidation_0-error:0.001483\tvalidation_1-error:0.00173\n",
      "[58]\tvalidation_0-error:0.001483\tvalidation_1-error:0.00173\n",
      "[59]\tvalidation_0-error:0.001483\tvalidation_1-error:0.001153\n",
      "[60]\tvalidation_0-error:0.001483\tvalidation_1-error:0.00173\n",
      "[61]\tvalidation_0-error:0.001483\tvalidation_1-error:0.00173\n",
      "[62]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[63]\tvalidation_0-error:0.001236\tvalidation_1-error:0.001153\n",
      "[64]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[65]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[66]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[67]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[68]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[69]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[70]\tvalidation_0-error:0.001236\tvalidation_1-error:0.001153\n",
      "[71]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[72]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[73]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[74]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[75]\tvalidation_0-error:0.001236\tvalidation_1-error:0.001153\n",
      "[76]\tvalidation_0-error:0.001236\tvalidation_1-error:0.00173\n",
      "[77]\tvalidation_0-error:0.001236\tvalidation_1-error:0.001153\n",
      "[78]\tvalidation_0-error:0.001236\tvalidation_1-error:0.001153\n",
      "[79]\tvalidation_0-error:0.001236\tvalidation_1-error:0.001153\n",
      "[80]\tvalidation_0-error:0.001236\tvalidation_1-error:0.001153\n",
      "[81]\tvalidation_0-error:0.001236\tvalidation_1-error:0.001153\n",
      "[82]\tvalidation_0-error:0.001236\tvalidation_1-error:0.001153\n",
      "[83]\tvalidation_0-error:0.000989\tvalidation_1-error:0.001153\n",
      "[84]\tvalidation_0-error:0.000989\tvalidation_1-error:0.001153\n",
      "[85]\tvalidation_0-error:0.000741\tvalidation_1-error:0.001153\n",
      "[86]\tvalidation_0-error:0.000741\tvalidation_1-error:0.001153\n",
      "[87]\tvalidation_0-error:0.000741\tvalidation_1-error:0.001153\n",
      "[88]\tvalidation_0-error:0.000741\tvalidation_1-error:0.001153\n",
      "[89]\tvalidation_0-error:0.000741\tvalidation_1-error:0.001153\n",
      "[90]\tvalidation_0-error:0.000741\tvalidation_1-error:0.00173\n",
      "[91]\tvalidation_0-error:0.000741\tvalidation_1-error:0.00173\n",
      "[92]\tvalidation_0-error:0.000741\tvalidation_1-error:0.00173\n",
      "[93]\tvalidation_0-error:0.000741\tvalidation_1-error:0.00173\n",
      "[94]\tvalidation_0-error:0.000741\tvalidation_1-error:0.001153\n",
      "[95]\tvalidation_0-error:0.000247\tvalidation_1-error:0.00173\n",
      "[96]\tvalidation_0-error:0.000247\tvalidation_1-error:0.00173\n",
      "[97]\tvalidation_0-error:0.000247\tvalidation_1-error:0.00173\n",
      "[98]\tvalidation_0-error:0.000247\tvalidation_1-error:0.00173\n",
      "[99]\tvalidation_0-error:0.000247\tvalidation_1-error:0.00173\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "XGBClassifier(base_score=0.5, colsample_bylevel=1, colsample_bytree=1,\n",
       "       gamma=0, learning_rate=0.1, max_delta_step=0, max_depth=3,\n",
       "       min_child_weight=1, missing=None, n_estimators=100, nthread=-1,\n",
       "       objective='binary:logistic', reg_alpha=0, reg_lambda=1,\n",
       "       scale_pos_weight=1, seed=0, silent=True, subsample=1)"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train and evaluate.\n",
    "xgb_model.fit(xtrain, ytrain, eval_metric=['error'], eval_set=[((xtrain, ytrain)),(xtest, ytest)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "confusion matrix\n",
      "[[1703    0]\n",
      " [   3   28]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00      1703\n",
      "          1       1.00      0.90      0.95        31\n",
      "\n",
      "avg / total       1.00      1.00      1.00      1734\n",
      "\n",
      "-----------------------------------------------------------------------------------------\n",
      "Accuracy is :\n",
      "0.998269896194\n",
      "Area under the curve : 0.951613\n"
     ]
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "y_pred = xgb_model.predict(xtest)  #predicts\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, y_pred))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, y_pred))\n",
    "print(\"-----------------------------------------------------------------------------------------\")\n",
    "print(\"Accuracy is :\")\n",
    "print(metrics.accuracy_score(ytest, y_pred))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, y_pred)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Higher F1 score comapred to Gradient Boosting Model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 12. SMOTE ( Synthetic Minority Over-sampling Technique)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting imblearn\n",
      "  Downloading https://files.pythonhosted.org/packages/81/a7/4179e6ebfd654bd0eac0b9c06125b8b4c96a9d0a8ff9e9507eb2a26d2d7e/imblearn-0.0-py2.py3-none-any.whl\n",
      "Collecting imbalanced-learn (from imblearn)\n",
      "  Downloading https://files.pythonhosted.org/packages/80/a4/900463a3c0af082aed9c5a43f4ec317a9469710c5ef80496c9abc26ed0ca/imbalanced_learn-0.3.3-py3-none-any.whl (144kB)\n",
      "\u001b[K    100% |████████████████████████████████| 153kB 5.9MB/s eta 0:00:01\n",
      "\u001b[?25hRequirement not upgraded as not directly required: numpy in /opt/conda/envs/DSX-Python35/lib/python3.5/site-packages (from imbalanced-learn->imblearn)\n",
      "Requirement not upgraded as not directly required: scikit-learn in /opt/conda/envs/DSX-Python35/lib/python3.5/site-packages (from imbalanced-learn->imblearn)\n",
      "Requirement not upgraded as not directly required: scipy in /opt/conda/envs/DSX-Python35/lib/python3.5/site-packages (from imbalanced-learn->imblearn)\n",
      "Installing collected packages: imbalanced-learn, imblearn\n",
      "Successfully installed imbalanced-learn-0.3.3 imblearn-0.0\n"
     ]
    }
   ],
   "source": [
    "!pip install imblearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Resampled dataset shape Counter({0: 56772, 1: 500})\n"
     ]
    }
   ],
   "source": [
    "from collections import Counter\n",
    "from imblearn.over_sampling import SMOTE \n",
    "\n",
    "x = df[['Time', 'V1', 'V2', 'V3', 'V4', 'V5', 'V6', 'V7', 'V8', 'V9', 'V10',\n",
    "       'V11', 'V12', 'V13', 'V14', 'V15', 'V16', 'V17', 'V18', 'V19', 'V20',\n",
    "       'V21', 'V22', 'V23', 'V24', 'V25', 'V26', 'V27', 'V28', 'Amount']]\n",
    "y = df['Class']\n",
    "\n",
    "'''Increase the fraud samples from 102 to 500'''\n",
    "\n",
    "sm = SMOTE(random_state=42,sampling_strategy={1:500})\n",
    "X_res, y_res = sm.fit_sample(x, y)\n",
    "print('Resampled dataset shape {}'.format(Counter(y_res)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "xtrain shape\n",
      "(40090, 30)\n",
      "xtest shape\n",
      "(17182, 30)\n"
     ]
    }
   ],
   "source": [
    "'''Split the resampled data into train & test data with 70:30 mix'''\n",
    "\n",
    "xtrain, xtest, ytrain, ytest = train_test_split(X_res, y_res, test_size=0.30, random_state=0)\n",
    "print('xtrain shape')\n",
    "print(xtrain.shape)\n",
    "print('xtest shape')\n",
    "print(xtest.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model\n",
      "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
      "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=1, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1,\n",
      "            oob_score=False, random_state=None, verbose=0,\n",
      "            warm_start=False)\n",
      "confusion matrix\n",
      "[[17019     4]\n",
      " [   14   145]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00     17023\n",
      "          1       0.97      0.91      0.94       159\n",
      "\n",
      "avg / total       1.00      1.00      1.00     17182\n",
      "\n",
      "Accuracy : 0.998952\n",
      "Area under the curve : 0.955857\n"
     ]
    }
   ],
   "source": [
    "'''Random Forest Classifier on resampled data'''\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn import metrics\n",
    "\n",
    "rfmodel = RandomForestClassifier()\n",
    "rfmodel.fit(xtrain,ytrain)\n",
    "print('model')\n",
    "print(rfmodel)\n",
    "\n",
    "ypredrf = rfmodel.predict(xtest)\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, ypredrf))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, ypredrf))\n",
    "print('Accuracy : %f' % (metrics.accuracy_score(ytest, ypredrf)))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, ypredrf)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### We can see an increase in F1 score and the Area under the curve. This is the advantage of synthetic over sampling technique which has given a good lift in accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "confusion matrix\n",
      "[[17011    12]\n",
      " [   16   143]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00     17023\n",
      "          1       0.92      0.90      0.91       159\n",
      "\n",
      "avg / total       1.00      1.00      1.00     17182\n",
      "\n",
      "-----------------------------------------------------------------------------------------\n",
      "Accuracy is :\n",
      "0.998370387615\n",
      "Area under the curve : 0.949333\n"
     ]
    }
   ],
   "source": [
    "'''Gradient Boost Algorithm on resampled data'''\n",
    "\n",
    "from sklearn import ensemble\n",
    "\n",
    "params = {'n_estimators': 500, 'max_depth': 3, 'subsample': 0.5,\n",
    "          'learning_rate': 0.01, 'min_samples_leaf': 1, 'random_state': 3}\n",
    "clf = ensemble.GradientBoostingClassifier(**params)\n",
    "clf.fit(xtrain, ytrain) #trains\n",
    "y_pred = clf.predict(xtest)  #predicts\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, y_pred))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, y_pred))\n",
    "print(\"-----------------------------------------------------------------------------------------\")\n",
    "print(\"Accuracy is :\")\n",
    "print(metrics.accuracy_score(ytest, y_pred))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, y_pred)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### There's an increase in precision, F1 score & Area under the curve using SMOTE and resampling the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'reg_alpha': 0, 'reg_lambda': 1, 'max_depth': 3, 'scale_pos_weight': 1, 'n_estimators': 100, 'max_delta_step': 0, 'min_child_weight': 1, 'seed': 0, 'colsample_bylevel': 1, 'colsample_bytree': 1, 'silent': 1, 'subsample': 1, 'learning_rate': 0.1, 'objective': 'binary:logistic', 'base_score': 0.5, 'gamma': 0, 'missing': None}\n"
     ]
    }
   ],
   "source": [
    "'''Extreme Gradient Boost Algorithm on reduced dimensions data'''\n",
    "\n",
    "from xgboost.sklearn import XGBClassifier\n",
    "\n",
    "# Create the XGB classifier, xgb_model.\n",
    "xgb_model = XGBClassifier()\n",
    "\n",
    "# List the default parameters.\n",
    "print(xgb_model.get_xgb_params())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\tvalidation_0-error:0.001197\tvalidation_1-error:0.001397\n",
      "[1]\tvalidation_0-error:0.001147\tvalidation_1-error:0.001455\n",
      "[2]\tvalidation_0-error:0.001147\tvalidation_1-error:0.001455\n",
      "[3]\tvalidation_0-error:0.001122\tvalidation_1-error:0.001455\n",
      "[4]\tvalidation_0-error:0.001147\tvalidation_1-error:0.001455\n",
      "[5]\tvalidation_0-error:0.001172\tvalidation_1-error:0.001455\n",
      "[6]\tvalidation_0-error:0.001172\tvalidation_1-error:0.001455\n",
      "[7]\tvalidation_0-error:0.001172\tvalidation_1-error:0.001455\n",
      "[8]\tvalidation_0-error:0.001172\tvalidation_1-error:0.001455\n",
      "[9]\tvalidation_0-error:0.001172\tvalidation_1-error:0.001513\n",
      "[10]\tvalidation_0-error:0.001172\tvalidation_1-error:0.001455\n",
      "[11]\tvalidation_0-error:0.001197\tvalidation_1-error:0.001513\n",
      "[12]\tvalidation_0-error:0.001197\tvalidation_1-error:0.001455\n",
      "[13]\tvalidation_0-error:0.001172\tvalidation_1-error:0.001455\n",
      "[14]\tvalidation_0-error:0.001122\tvalidation_1-error:0.001455\n",
      "[15]\tvalidation_0-error:0.001048\tvalidation_1-error:0.001455\n",
      "[16]\tvalidation_0-error:0.001048\tvalidation_1-error:0.001455\n",
      "[17]\tvalidation_0-error:0.001023\tvalidation_1-error:0.001455\n",
      "[18]\tvalidation_0-error:0.001023\tvalidation_1-error:0.001455\n",
      "[19]\tvalidation_0-error:0.001023\tvalidation_1-error:0.001455\n",
      "[20]\tvalidation_0-error:0.001048\tvalidation_1-error:0.001513\n",
      "[21]\tvalidation_0-error:0.001048\tvalidation_1-error:0.001455\n",
      "[22]\tvalidation_0-error:0.001073\tvalidation_1-error:0.001513\n",
      "[23]\tvalidation_0-error:0.001098\tvalidation_1-error:0.001513\n",
      "[24]\tvalidation_0-error:0.001023\tvalidation_1-error:0.001455\n",
      "[25]\tvalidation_0-error:0.001023\tvalidation_1-error:0.001513\n",
      "[26]\tvalidation_0-error:0.000973\tvalidation_1-error:0.001455\n",
      "[27]\tvalidation_0-error:0.000973\tvalidation_1-error:0.001455\n",
      "[28]\tvalidation_0-error:0.000973\tvalidation_1-error:0.001455\n",
      "[29]\tvalidation_0-error:0.000973\tvalidation_1-error:0.001397\n",
      "[30]\tvalidation_0-error:0.000973\tvalidation_1-error:0.001339\n",
      "[31]\tvalidation_0-error:0.000948\tvalidation_1-error:0.00128\n",
      "[32]\tvalidation_0-error:0.000948\tvalidation_1-error:0.00128\n",
      "[33]\tvalidation_0-error:0.000923\tvalidation_1-error:0.00128\n",
      "[34]\tvalidation_0-error:0.000923\tvalidation_1-error:0.00128\n",
      "[35]\tvalidation_0-error:0.000923\tvalidation_1-error:0.00128\n",
      "[36]\tvalidation_0-error:0.000973\tvalidation_1-error:0.00128\n",
      "[37]\tvalidation_0-error:0.000998\tvalidation_1-error:0.00128\n",
      "[38]\tvalidation_0-error:0.000973\tvalidation_1-error:0.00128\n",
      "[39]\tvalidation_0-error:0.000973\tvalidation_1-error:0.00128\n",
      "[40]\tvalidation_0-error:0.000973\tvalidation_1-error:0.001222\n",
      "[41]\tvalidation_0-error:0.000948\tvalidation_1-error:0.001222\n",
      "[42]\tvalidation_0-error:0.000973\tvalidation_1-error:0.001222\n",
      "[43]\tvalidation_0-error:0.000948\tvalidation_1-error:0.001222\n",
      "[44]\tvalidation_0-error:0.000948\tvalidation_1-error:0.001222\n",
      "[45]\tvalidation_0-error:0.000948\tvalidation_1-error:0.001164\n",
      "[46]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001222\n",
      "[47]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001164\n",
      "[48]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001164\n",
      "[49]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001164\n",
      "[50]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001222\n",
      "[51]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001222\n",
      "[52]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001222\n",
      "[53]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001164\n",
      "[54]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001164\n",
      "[55]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001164\n",
      "[56]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001164\n",
      "[57]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001164\n",
      "[58]\tvalidation_0-error:0.000873\tvalidation_1-error:0.001164\n",
      "[59]\tvalidation_0-error:0.000848\tvalidation_1-error:0.001164\n",
      "[60]\tvalidation_0-error:0.000848\tvalidation_1-error:0.001164\n",
      "[61]\tvalidation_0-error:0.000848\tvalidation_1-error:0.001164\n",
      "[62]\tvalidation_0-error:0.000848\tvalidation_1-error:0.001164\n",
      "[63]\tvalidation_0-error:0.000848\tvalidation_1-error:0.001164\n",
      "[64]\tvalidation_0-error:0.000848\tvalidation_1-error:0.001164\n",
      "[65]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001164\n",
      "[66]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001164\n",
      "[67]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001106\n",
      "[68]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001164\n",
      "[69]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001164\n",
      "[70]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001164\n",
      "[71]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001106\n",
      "[72]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001106\n",
      "[73]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001106\n",
      "[74]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001048\n",
      "[75]\tvalidation_0-error:0.000823\tvalidation_1-error:0.001048\n",
      "[76]\tvalidation_0-error:0.000773\tvalidation_1-error:0.001048\n",
      "[77]\tvalidation_0-error:0.000773\tvalidation_1-error:0.001048\n",
      "[78]\tvalidation_0-error:0.000773\tvalidation_1-error:0.001048\n",
      "[79]\tvalidation_0-error:0.000773\tvalidation_1-error:0.001048\n",
      "[80]\tvalidation_0-error:0.000773\tvalidation_1-error:0.001048\n",
      "[81]\tvalidation_0-error:0.000773\tvalidation_1-error:0.001048\n",
      "[82]\tvalidation_0-error:0.000748\tvalidation_1-error:0.001048\n",
      "[83]\tvalidation_0-error:0.000748\tvalidation_1-error:0.001106\n",
      "[84]\tvalidation_0-error:0.000748\tvalidation_1-error:0.001106\n",
      "[85]\tvalidation_0-error:0.000748\tvalidation_1-error:0.001106\n",
      "[86]\tvalidation_0-error:0.000748\tvalidation_1-error:0.000989\n",
      "[87]\tvalidation_0-error:0.000723\tvalidation_1-error:0.001048\n",
      "[88]\tvalidation_0-error:0.000698\tvalidation_1-error:0.000989\n",
      "[89]\tvalidation_0-error:0.000673\tvalidation_1-error:0.000931\n",
      "[90]\tvalidation_0-error:0.000673\tvalidation_1-error:0.000931\n",
      "[91]\tvalidation_0-error:0.000624\tvalidation_1-error:0.000931\n",
      "[92]\tvalidation_0-error:0.000549\tvalidation_1-error:0.000873\n",
      "[93]\tvalidation_0-error:0.000549\tvalidation_1-error:0.000873\n",
      "[94]\tvalidation_0-error:0.000549\tvalidation_1-error:0.000931\n",
      "[95]\tvalidation_0-error:0.000549\tvalidation_1-error:0.000931\n",
      "[96]\tvalidation_0-error:0.000524\tvalidation_1-error:0.000931\n",
      "[97]\tvalidation_0-error:0.000499\tvalidation_1-error:0.000873\n",
      "[98]\tvalidation_0-error:0.000499\tvalidation_1-error:0.000873\n",
      "[99]\tvalidation_0-error:0.000499\tvalidation_1-error:0.000873\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "XGBClassifier(base_score=0.5, colsample_bylevel=1, colsample_bytree=1,\n",
       "       gamma=0, learning_rate=0.1, max_delta_step=0, max_depth=3,\n",
       "       min_child_weight=1, missing=None, n_estimators=100, nthread=-1,\n",
       "       objective='binary:logistic', reg_alpha=0, reg_lambda=1,\n",
       "       scale_pos_weight=1, seed=0, silent=True, subsample=1)"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train and evaluate.\n",
    "xgb_model.fit(xtrain, ytrain, eval_metric=['error'], eval_set=[((xtrain, ytrain)),(xtest, ytest)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "confusion matrix\n",
      "[[17019     4]\n",
      " [   11   148]]\n",
      "classification report\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       1.00      1.00      1.00     17023\n",
      "          1       0.97      0.93      0.95       159\n",
      "\n",
      "avg / total       1.00      1.00      1.00     17182\n",
      "\n",
      "-----------------------------------------------------------------------------------------\n",
      "Accuracy is :\n",
      "0.999126993365\n",
      "Area under the curve : 0.965291\n"
     ]
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "y_pred = xgb_model.predict(xtest)  #predicts\n",
    "print('confusion matrix')\n",
    "print(metrics.confusion_matrix(ytest, y_pred))\n",
    "print('classification report')\n",
    "print(metrics.classification_report(ytest, y_pred))\n",
    "print(\"-----------------------------------------------------------------------------------------\")\n",
    "print(\"Accuracy is :\")\n",
    "print(metrics.accuracy_score(ytest, y_pred))\n",
    "print('Area under the curve : %f' % (metrics.roc_auc_score(ytest, y_pred)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### XGBoost has given high F1 score when compared with Gradient Boosting Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### We can conclude stating that sampling techniques are important and can be helpful if the data is skewed. We have seen two statistical algorithms (Bagging & Boosting) and their implementation on different samples of data providing different outcomes. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}