clustering_parte3.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Clusterização\n",
    "\n",
    "## Trabalhando com dados categóricos\n",
    "\n"
   ]
  },
  {
   "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>idade</th>\n",
       "      <th>conta_corrente</th>\n",
       "      <th>risco</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>21.257389</td>\n",
       "      <td>783.127911</td>\n",
       "      <td>ruim</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>21.766573</td>\n",
       "      <td>979.747521</td>\n",
       "      <td>bom</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>25.321033</td>\n",
       "      <td>1065.328054</td>\n",
       "      <td>ruim</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>23.919268</td>\n",
       "      <td>1195.758078</td>\n",
       "      <td>bom</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>21.805298</td>\n",
       "      <td>1083.764450</td>\n",
       "      <td>bom</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       idade  conta_corrente risco\n",
       "0  21.257389      783.127911  ruim\n",
       "1  21.766573      979.747521   bom\n",
       "2  25.321033     1065.328054  ruim\n",
       "3  23.919268     1195.758078   bom\n",
       "4  21.805298     1083.764450   bom"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "df = pd.read_csv('dados/exemplo2.csv')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Nesse exemplo, temos uma coluna indicando o risco do cliente de forma categórica: bom ou ruim. Para a clusterização, precisamos transformar dados categóricos em dados numéricos.\n",
    "\n",
    "O forma mais adequado para a quantificação de dados categóricos é através do método conhecido como _one hot encoder_. Esse método utiliza a binarização para a tranformação dos dados. No ```exemplo2```, temos duas opções de valores para a coluna ```risco```: bom ou ruim. Para binarizar essa coluna, criaremos uma nova coluna identificada como ```risco_bom```, tendo valor igual a 1, caso o risco seja bom e 0 caso contrário. \n",
    "\n",
    "Para executar a transformação criaremos uma função chamada ```encoder()```, que recebe a coluna (nesse caso o risco) e a categoria que corresponde ao valor 1 (verdadeiro), retornando 1, caso ```coluna = categoria``` e 0, caso contrário."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def encoder(column, true_value):\n",
    "    if column == true_value:\n",
    "        return 1.0\n",
    "    else:\n",
    "        return 0.0"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para criar a nova coluna com base na função definida acima, utilizamos o comando ```Series.apply(função, args=())```."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>idade</th>\n",
       "      <th>conta_corrente</th>\n",
       "      <th>risco</th>\n",
       "      <th>risco_bom</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>21.257389</td>\n",
       "      <td>783.127911</td>\n",
       "      <td>ruim</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>21.766573</td>\n",
       "      <td>979.747521</td>\n",
       "      <td>bom</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>25.321033</td>\n",
       "      <td>1065.328054</td>\n",
       "      <td>ruim</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>23.919268</td>\n",
       "      <td>1195.758078</td>\n",
       "      <td>bom</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>21.805298</td>\n",
       "      <td>1083.764450</td>\n",
       "      <td>bom</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       idade  conta_corrente risco  risco_bom\n",
       "0  21.257389      783.127911  ruim        0.0\n",
       "1  21.766573      979.747521   bom        1.0\n",
       "2  25.321033     1065.328054  ruim        0.0\n",
       "3  23.919268     1195.758078   bom        1.0\n",
       "4  21.805298     1083.764450   bom        1.0"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['risco_bom'] = df.risco.apply(encoder, args=('bom',))\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Eliminaremos, agora, o coluna ```risco``` utilizando o comando ```DataFrame.drop(\"nome_da_coluna\", axis=1, inplace=True)```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>idade</th>\n",
       "      <th>conta_corrente</th>\n",
       "      <th>risco_bom</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>21.257389</td>\n",
       "      <td>783.127911</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>21.766573</td>\n",
       "      <td>979.747521</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>25.321033</td>\n",
       "      <td>1065.328054</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>23.919268</td>\n",
       "      <td>1195.758078</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>21.805298</td>\n",
       "      <td>1083.764450</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       idade  conta_corrente  risco_bom\n",
       "0  21.257389      783.127911        0.0\n",
       "1  21.766573      979.747521        1.0\n",
       "2  25.321033     1065.328054        0.0\n",
       "3  23.919268     1195.758078        1.0\n",
       "4  21.805298     1083.764450        1.0"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.drop('risco', axis=1, inplace=True)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Com a nova coluna, podemos aplicar algoritmos de clusterização. Faremos um teste com o ```Kmeans```, lembrando de normalizar os dados utilizando o ```MinMaxScaler```.\n",
    "\n",
    "__Observação__: como a coluna ```risco_bom``` já está definida na escala entre 0 e 1, a mesma não será afetada pela normalização."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.cluster import KMeans\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "\n",
    "X = MinMaxScaler().fit_transform(df[['idade', 'conta_corrente', 'risco_bom']])\n",
    "df['km_3'] = KMeans(n_clusters=3, random_state=1).fit_predict(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "def plot(x, y, column_name):\n",
    "    plt.figure(figsize=(10, 7))\n",
    "    plt.xlabel(x)\n",
    "    plt.ylabel(y)\n",
    "    for c in sorted(df[column_name].unique()):\n",
    "        plt.scatter(df[df[column_name] == c][x], df[df[column_name] == c][y]) \n",
    "    plt.legend([f'cluster {c}' for c in sorted(df[column_name].unique())])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmoAAAGpCAYAAAA9Rhr4AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nOy9e3iT55nue7862JJNbBEgsTDllFCSDRhMCOGQkNNsu1O3wKKkpMlKWDOZds/OrImnaZnKOeEkUNySaUt316S70zaLZHU1EMKCtJ5d3CmZMZBFGIgdUyZQiJM2NiYQgmywZVuHd/8hfeKT9L7fQWfJz++6uLA+fYdXki3deg73wzjnIAiCIAiCIPIPS64XQBAEQRAEQYghoUYQBEEQBJGnkFAjCIIgCILIU0ioEQRBEARB5Ckk1AiCIAiCIPIUW64XkCkmTpzIp0+fnutlEARBEARB6HLs2LFPOOeT4rcXrVCbPn06jh49mutlEARBEARB6MIY+6NoO6U+CYIgCIIg8hQSagRBEARBEHkKCTWCIAiCIIg8pWhr1AiCIAiCyAx+vx89PT0YHh7O9VIKDofDgSlTpsButxvan4QaQRAEQRCm6OnpwTXXXIPp06eDMZbr5RQMnHNcvHgRPT09mDFjhqFjKPVJEARBEIQphoeHMWHCBBJpJmGMYcKECaYikSTUCIIgCIIwDYm05DD7vJFQIwiCIAiCyFNIqBEEQRAEURQ0NzfjhRdeMH2c1+vFP/7jP6Z8/Q8++AC33XYbZs2ahXXr1mF0dDTlc5JQIwiCIAhiTJOMUOOcIxQKxWz71re+ha9//es4ffo0xo8fj5/97Gcpr42EGkEQBEEQGWVPRy+Wt+zHDE8rlrfsx56O3pTP+fLLL6Ompgbz58/HQw89lHD/XXfdFR0l+cknn0CZ/33ixAksXrwYCxYsQE1NDU6fPg2Px4P3338fCxYswIYNGwAAW7duxa233oqamhps3LgRAPDhhx/i5ptvxqOPPoqFCxfio48+il6Pc479+/dj7dq1AID169djz549KT9OsucgCIIgCCJj7OnoRdPu4/D5gwCAXq8PTbuPAwBW11Yndc4TJ05g8+bNOHToECZOnIhPP/3U8LE//vGP0djYiAcffBCjo6MIBoNoaWnB73//e3R2dgIA2tracPr0aRw5cgScc6xcuRLt7e2YOnUqTp06hZdeeikhAnfx4kW4XC7YbGFpNWXKFPT2pi5ISagRBEEQBJExtu47FRVpCj5/EFv3nUpaqCmRq4kTJwIArr32WsPHLl26FJs3b0ZPTw/WrFmDWbNmJezT1taGtrY21NbWAgCuXLmC06dPY+rUqZg2bRqWLFmScAznPGFbOjpjKfVJEARBEETGOOv1mdpuBM65rgiy2WzRGjK1b9kDDzyAN954A06nE/X19di/f7/w/E1NTejs7ERnZyfOnDmDRx55BABQXl4uvN7EiRPh9XoRCAQAhE2BJ0+enNTjU0NCjSAIgiCIjDHZ5TS13Qj33nsvdu7ciYsXLwKAMPU5ffp0HDt2DACwa9eu6Pbu7m7MnDkTjz32GFauXImuri5cc801uHz5cnSf+vp6/PznP8eVK1cAAL29vTh//rzmmhhjuPvuu6PX2r59O1atWpX0Y1QgoUYQOaa1uxV1u+pQs70Gdbvq0NrdmuslEQRBpI0N9bPhtFtjtjntVmyon530OefMmYMnn3wSd955J+bPn4/HH388YZ9vfvObePHFF7Fs2TJ88skn0e07duzA3LlzsWDBApw8eRIPP/wwJkyYgOXLl2Pu3LnYsGED6urq8MADD2Dp0qWYN28e1q5dGyPkZHznO9/B9773Pdx44424ePFiNAqXCkyUUy0GFi1axJVuD4LIV1q7W9H8VjOGg1fD8g6rA83LmtEwsyGHKyMIgpDz3nvv4eabbza8/56OXmzddwpnvT5MdjmxoX520vVpxYDo+WOMHeOcL4rfl5oJCCKHbHtnW4xIA4Dh4DC2vbONhBpBEEXD6trqMS3MUoFSnwSRQ84NnjO1nSAIghhbkFAjiBxSVV5lajtBEAQxtiChRhA5pHFhIxxWR8w2h9WBxoWNOVoRQRAEkU9QjRpB5BClDm3bO9twbvAcqsqr0LiwkerTCIIgCAAk1Agi5zTMbCBhRhAEQQih1CdBFDHk0UYQxFiiubkZL7zwgunjvF5vwuzOZPjRj36EG2+8EYyxGO+2VCChRhBFiuLR1jfYBw6OvsE+NL/VbEiskcAjCGIskYxQ45xHR1QpLF++HP/yL/+CadOmpW1tJNQIokjR8mjTIhWBRxAEIaRrJ/D9uUCzK/x/186UT/nyyy+jpqYG8+fPx0MPPZRw/1133QXF+P6TTz7B9OnTAQAnTpzA4sWLsWDBAtTU1OD06dPweDx4//33sWDBAmzYsAEAsHXrVtx6662oqanBxo0bAQAffvghbr75Zjz66KNYuHAhPvroo5hr1tbWRq+TLqhGjSCKlGQ92siElyCItNK1E/jVY4A/MoS9/6PwbQCo+XJSpzxx4gQ2b96MQ4cOYeLEicJZnzJ+/OMfo7GxEQ8++CBGR0cRDAbR0tKC3//+9+js7AQAtLW14fTp0zhy5Ag451i5ciXa29sxdepUnDp1Ci+99FJaUqVGIKFGEEVKVXkV+gb7hNu1MCLwaBwMQRCG+d1zV0Wagt8X3p6kUNu/fz/Wrl2LiRMnAgCuvfZaw8cuXboUmzdvRk9PD9asWYNZs2Yl7NPW1oa2tjbU1tYCAK5cuYLTp09j6tSpmDZtGpYsWZLUupOBUp8EUaSsmLIiYZsRjzY9E949Hb1o2n0cvV4fOIBerw9Nu49jT0dvymsmCKII6e8xt90AnHMwxjT3sdls0Rqy4eGrWYIHHngAb7zxBpxOJ+rr67F//37h+ZuamtDZ2YnOzk6cOXMmOmC9vLw86XUnAwk1gihCWrtbsffM3oTtq25cpZu+1DPh3brvFHz+YMz9Pn8QW/edSnHVBEEUJZVTzG03wL333oudO3fi4sWLACBMfU6fPh3Hjh0DAOzatSu6vbu7GzNnzsRjjz2GlStXoqurC9dccw0uX74c3ae+vh4///nPceXKFQBAb28vzp8/n/R6U4GEGpEU1BWY34jqzACgvadd99iGmQ1oXtYMd7kbDAzucjealzVHBd5Zr094nGw7QRBjnHufAezO2G12Z3h7ksyZMwdPPvkk7rzzTsyfPx+PP/54wj7f/OY38eKLL2LZsmUxVhk7duzA3LlzsWDBApw8eRIPP/wwJkyYgOXLl2Pu3LnYsGED6urq8MADD2Dp0qWYN28e1q5dGyPkZPzwhz/ElClT0NPTg5qaGvzVX/1V0o9RgXHOUz5JPrJo0SKudHsQ6UXpClQLAYfVEfNhTuSWmu014Ej822Zg6FrfldK5l7fsR69AlFW7nDjkuSelcxMEURi89957uPnmm40f0LUzXJPW3xOOpN37TNL1acWA6PljjB3jnC+K35ciaoRpkrV9IJIjmehlJoe9b6ifDafdGrPNabdiQ/3slM9NEESRUvNl4Ou/B5q94f/HsEgzCwk1wjTJ2j4Q5knW0yyTw95X11Zjy5p5qHY5wRCOpG1ZM4+6PgmCIDIA2XMQpknW9qFQae1uzdnQ9GQ9zTI97H11bTUJM4IgiCxAQo0wTePCRmGNWjqiNSIyKZT0zh1fj6dEtABkRayJBDFgLHpJw94JgiAKH0p9EqbR6wpMJ5kcZ2Tk3Lmsx9N6jBUlFabPlUyXLnX3EgRB5BaKqBFJka1oTSbHGRk5dy7r8bTEoJ7Ro5pko4LZjibmMsVMEASRr1BEjchrMimUjJw7k92Temg9xv6RfsPnSTYqmM1oIg2CJwgiHTQ3N+OFF14wfZzX603L7M4HH3wQs2fPxty5c/GXf/mX8Pv9KZ+ThBqR12RSKBk5dya7J/XQeoxV5VUJaclNhzcJ05TJit1sRhPJ8oUgiFySjFDjnEdHVCk8+OCDOHnyJI4fPw6fz4ef/vSnKa+NhBqR12RSKBk5dzbr8UTrs1vsCdttzIYVU1YkRKB2nNohjEglK3azGU0kyxeCKG4yUe/68ssvo6amBvPnz8dDDz2UcP9dd90Fxfj+k08+wfTp0wEAJ06cwOLFi7FgwQLU1NTg9OnT8Hg8eP/997FgwQJs2LABALB161bceuutqKmpwcaNGwEAH374IW6++WY8+uijWLhwIT766KOYa37+858HYwyMMSxevBg9PcnPM1UgoUbkNZkUSupzA4CFWaJRHPWbSMPMBrStbcOWO7YAAJoONGWlsL5hZgOeX/48XKWu6LbKkkpsun0T2nvahSOi1CiPJVmxm81oYi5TzARBZJZMlDacOHECmzdvxv79+/Huu+9i2zbj0fcf//jHaGxsRGdnJ44ePYopU6agpaUFN9xwAzo7O7F161a0tbXh9OnTOHLkCDo7O3Hs2DG0t4dH8J06dQoPP/wwOjo6MG3aNOE1/H4/XnnlFXzuc59L+jEqZLSZgDHmAvBTAHMBcAB/CeAUgB0ApgP4EMCXOeeXWLg6ehuAzwMYAvBfOOfvRM6zHsBTkdNu4pxvz+S6ifwik40Lynn1iuZzZdMhe+xNB5oMHa/YezQvazZdqC/yYlsxZQW2vbMNTQea0lrwn23LF4IgskcmmsL279+PtWvXYuLEiQCAa6+91vCxS5cuxebNm9HT04M1a9Zg1qxZCfu0tbWhra0NtbW1AIArV67g9OnTmDp1KqZNm4YlS5ZoXuPRRx/FihUrcMcdd5h4VGIyHVHbBuA3nPObAMwH8B4AD4Dfcc5nAfhd5DYA/DmAWZF/XwPwIgAwxq4FsBHAbQAWA9jIGBuf4XUTYwgj9VH5VkNlJtKkCMq2tW3oWt+FtrVtht4c47swV0xZgb1n9mak4D+XKWaCIDJLJkobOOe63e82my1aQzY8fPX9+4EHHsAbb7wBp9OJ+vp67N+/X3j+pqYmdHZ2orOzE2fOnMEjjzwCACgvL9e87rPPPosLFy7ge9/7ntmHJSRjQo0xVgFgBYCfAQDnfJRz7gWwCoASEdsOYHXk51UAXuZhDgNwMcbcAOoB/JZz/inn/BKA3wJIPZZIEBGMvImk8kYjqs3YdHgT5r88H/O2z8P8l+dj0+FNptYsSkvKUKdzjdaItHa34qmDTyXUwGVSrCopZjNikiCI/CcTpQ333nsvdu7ciYsXLwIAPv3004R9pk+fjmPHjgEAdu3aFd3e3d2NmTNn4rHHHsPKlSvR1dWFa665BpcvX47uU19fj5///Oe4cuUKAKC3txfnz5/XXddPf/pT7Nu3D7/85S9hsaRHYmUyojYTwAUALzHGOhhjP2WMlQO4nnPeBwCR/6+L7F8NQF2V1xPZJtueAGPsa4yxo4yxoxcuXEjvoyGKFiNvIjKDWb03GlFthueABztO7UCIh7/phXgIO07tMCXW4iNQeijRL6PRsC1vb0GABwytJV6skkkuQRBqMlHvOmfOHDz55JO48847MX/+fDz++OMJ+3zzm9/Eiy++iGXLluGTTz6Jbt+xYwfmzp2LBQsW4OTJk3j44YcxYcIELF++HHPnzsWGDRtQV1eHBx54AEuXLsW8efOwdu3aGCEn46//+q/x8ccfY+nSpViwYAGee+65pB+jAuOcp3wS4YkZWwTgMIDlnPO3GWPbAAwA+FvOuUu13yXO+XjGWCuALZzzg5HtvwPw9wDuAVDKOd8U2f40gCHO+T9oXX/RokVc6fYgCC3i68+A8JuIknpr7W7F04eehj8U64djYzZsun2TZuSnbleddAxUPBZmwbsPv5vUY5i3fZ7uuRVhqMZd7kbb2jbT55OdQ++5JAiiOHjvvfdw8803G96fDK1jET1/jLFjnPNF8ftmspmgB0AP5/ztyO1dCNejfcwYc3PO+yKpzfOq/T+jOn4KgLOR7XfFbf/XDK6bMEEx/PHpDTDf9s62BJEGAAEeQMuRlphzxGOmBkMkpIziLndLBaHD6pB2iKZqfxH/rTiTkyQIgihcaPZw8mQs9ck5PwfgI8bY7MimewH8B4A3AKyPbFsPYG/k5zcAPMzCLAHQH0mN7gNQxxgbH2kiqItsI3JMttzk05FK0ztHw8wGNC5sRFV5Fc4Nnoux6NASM94RL54+9LR0TWZqMCws+T9HWc2aq9QVY0FidH1qSxA1TqtTs+Cf/NAIgiDSS6Znff4tgF8wxkoAdAP4C4TF4U7G2CMA/gTgvsi+/4ywNccZhO05/gIAOOefMsaeB/Dvkf2e45wnVg0SWScb0ZN02GIYOYfWPlXlVZrpS3/IL33MItsJGfd99j7dfWToRQUBCNfhC/jQ2t2asHbPYk9CutdusWPjso2az7vsuSI/NIIoPox0XhKJmC05y1iNWq6hGrXMU7O9BhyJvz8MDF3ru5I+b2t3K1qOtMA74pXuI6utEiGrE3OXu9G4sBHb3tkmFWLKPnpiS+sxq9PDFSUVGAoMJaRSl1QtwT/V/5Ohx6OHLB3d2t2KLW9vQf9o7JxQWQ2ZkbS2zMKDatQIorj54IMPcM0112DChAkk1kzAOcfFixdx+fJlzJgxI+Y+WY0aCTUiabQEULyIMlrLJivcj8eMGJQJSkC7fkt9ndbuVjxx8AlpHZkZ4ZhKXZ+WCJMJTrVQMvOaGVmLqHFg1Y2r0N7TnpG6xWKoiSSIYsDv96OnpyfGn4wwhsPhwJQpU2C3x44IJKFGpB2tDj8AmlEkWZTFaJdkfKehWfECyDshta7z1MGnEmwr7BY7nl/+fMYFg5Ywio9ixVNZUomDXzko7eZMJgqaTtFnBOooJQiimMlF1ydR5MjqooDYeqj4VBsgr2UzUnSu7jSU1ZZ1nO/QFC96kbT466gfrzp96Cp1wbPYkxWhIKsJfO0Pr+kKzv7Rfk2ftmRqyLLdOEAdpQRBjEVIqBEpIWq5rttVZ6h4XvSBrle4r9SMqUWiWfHCwKLpOa1omyhSk8sWc5kAMmrr8dofXpPel4zxZCYaBzYd3hR97SzMgvs+ex+eWhIe80sdpQRBjEUyPeuTGIMY/eAUfaBrCQZXqSthtFAy4oWDY/fp3VgxZYV8H85TFmTpduiXCSCjth5az0kyjzXdbuObDm/SnNiQiTE0BEEQ+Q4JNSLtGPnglH2gawmG/pHEFGqy4sUf8qO9px2VJZXC+1P98N90eBM8BzxJeczJBJ7MK23x9YsNzf2UPScyjzU90j1IXRbxU7ZnYgwNQRBEvkNCjUg7og9UG7PBVeoy9IFuxpxV9uF932fv0xUv5wbPoem2JumHf7IRsdbuVuw4tSNhu5EB5lomwg0zG7DqxlUJx3Re6MSqG1dFBZOr1AUbi61qkD0nIqGj9bjj7wOQtkHqsoifsj3dwpAgCKIQoBo1Iu1oma8qnZhNB5qw7Z1tCduVLlG7xZ7QJaoVgRNdq/a6Wk1LjaryKsMNEWaMdrXEmF5aWKtgHhBHnYaDw2jvaY/pTpU1PNReVxtzn8MWK9y0jH+B5J8TI8i6cNWRQBpDQxDEWIOEGpERRB+oWh2ar//h9ajtRf9ofzQy1D/Sr+uXJfvwVrbJLDVWTFmBul11UYG25Y4t0WNEDRHDwWE8cfCJmHOL0BJjeilV2bHKcyUTncpxIguL4UDs4xgJjkR/9o54Y8SWnlBMtetSywftvs/eJ4xEpjKxgSAIotAhoUZkDZkI2HlqZ4IhLQcH5zzG2ysZs1OZpUb99PoY+4746JBWk4Jo/JR6XRUlFUJLEkC/u1LWSWlhFs1OWkUA6llY6N2fTGel0eYRvVFeSnenrOuTIAhiLEKGt0RWaO1uheeAx/Rxx9cfjx6fTrNTLWNdC7PAYXVgKDAkPd7CLOCcC8187RY7OOcJUbx1s9fpig7R49RD/TzIDG2B8HOpN/ZLy8QWQEoGt9k2yCUIgigkZIa31ExAZBxFfKSCXkrO7Hq0vNpCPIShwBCszKq5DwdH/2h/wrgrf8iPcSXjYoreW+5oMRQZkjUMyJD5vYlY3rIfwVHtLleZZcmKKStS7rokHzSCIAjzUOqTyDgikaXgsDrAwOAL+oT31+2qQ+PCxqQ/5GVDw40Q5EFDY6ZEeEf6ceD+A6aPA4D2nnZD+5mJKHIOeCdsRGh0Apjdi/gZyopAk127vac9KjRFjRfqWj9ZSlqWEq4oqdBdP0EQxFiFhBqRcbTElDIXVDaIXaljkn3IaxXni2qiRMXqWnz79m+bTkUCAPdX4pFfNeHopX+O1lstvn4x/nj5jzg3eA6VpZXgnGNgdCBB3Gg9X+5yd1IDyRkDWIlXKNKAqwJNTxDHN27o1Z3FrkFwYUD4uhMEQRBhSKgRGUdWIK/UPW17Zxv8Ib80ejUcHIbD5kiYz2ljNvgCPtRsr4mJ7igRH8ZYUtEwBQuzoGFmA45++Cle/+CfELJeAoMFYNrn5CE7giMTcOTir4GINgnxEA73HY7e9o54o/vHi5vK0sqY+xWUyQwKiqeZWri5Sl3CYxUkWikqxGTXriytFDZzmJm/KTIsBoChwFDUJ44gCIKIhWrUCNOYNYIV1T05rA6smLIiau4KaI846h/px6obV8V4anFweEe8UWPYpw4+hacPPR01i01FpAFhW4g9Hb149c1JGPq4DtzvAkcI8f03PGRBKFAGzoHQqAvDfWtgLe+OirIoEpEExNbbyRp81Ntlxrj10+tht9hNP1YlMim79uCID55/ezrherJaP9kcVxnJ1BoSBEGMBSiiRuiijqQ4rI6YejI909PW7lZhTZgyFN1oSrGipAJ7z+yNEV9BHozZJ8ADEDQ0JoWVWVF7XS2a3/gFLNNfh8M6FBuN4gAYEAqUAeBg1iFwvwsjF+oRGKgFJptLsQJXOyoHRgeE96u3yyJZ7T3teH7589j2zjapiOI8NrJmt9gx5B+SdoQCwCgfQfwEquHgsDQKqoiyPR292LrvFM56fZhY9WfA+FeE55elXJOxZCEIgigmKKJGaBIfuREV/Wt1X8oaCdp72g13+zmsDjDGTNeJpUKQB7Hl7S3wVb4Ki20oMWXIwqlIi8UPi80HxgBLiRcO927YKjqgGT7ToG5XHSpL9eePakWyGmY2oG1tG1ruaEno0uQhO+yDy1Fpvy5qKsx5uHtVJtK0CPGQtBN0T0cvmnYfR6/XBw7gwrk54AGn7mNT0BqnRRAEMVYgoUZootWxqcZsV2bfYJ+0uNxV6kqY5yirb8ok/aP9YBZ5obt3xAvE3c8sfpRO2ofpJfcmFd3rG+yT1ph5h72o2V6D2395u/R4teARzcb8zp3Po+NvfoyDD/wOXeu74LQ5E/ze4omPwKlRXh/R/M2t+07B54+Neg5/vBIIxaZmZRYf6bRkIQiCKFQo9UloYjTqJas/kjUSAOKaNIfVgfrp9Qk2EVpu//mGpaQfv37g+5rms8mgRDO1nod4waPu0lTPWVXSiHqvr6Y9CUc0FSlKR571JkZfAwO1GAYw47Pt6Bvsi05cUMSX+jzku0YQBEERNUIHvdmUCjLTU5FJqgwLs2DVjauw98zemHTXUwefwhX/FcNrTheuUpf0PofVgcoScYrSXV6FPR290jSfnZkv9jeKrH5LlkbU8zDjnKOy5DrhfWW2Cs16scku8eO/zrIs+nuhiEBRWlNL/BMEQYwVSKgRmhgRWutmr9McmG7UaZ9zLmwwCPBAQuMAEB57lCmcVqe0A7KypBLNy5rRdFuTtD5r675TGP54JXgo7k+MA+Ul5RlZs2J3oqDuzn3i4BPCNCJjDDYmD6xXlVehacnjsLPSmO12Vopnlj2huZ4N9bPhtMdOd3DardhQP9tQWjPVSQgEQRDFAKU+CU0UAfbEwSeEKbDKkkrd0UhGnfaryqtMpbU4eIK3WroYCY4kNE64Sl3wLPYkiFL1wHeHLSwsznp9sFYAPOQAWHhmKGMAGDR9zpIlXsDEG9HKRGf/SL/UOw24mtrsON8RMyx9zWdXo2Fmg2ZX5uraagCIdn1OdjmxoX42VtdW45ku/bSmch69rk/qDCUIopihoeyEIfSGeSdzbDzrZq9De0+75hxONe5yN1ZMWREVENlAqdlyl7vRuLARHec7EqYdOKwOBPpvgb/siGYzQjppuaMlRpxoDZ1Xo0w6kL0+x9cfFw6Kd1gd0TR1/HYjY61k66ssqcTBrxzUXbeCbG1GR2sRBEHkCzSUnUiJVOqFjNYUtfe0C9NdNmZLMHG1W+zwDnux49SOrIk0ADE1VU8dfEo4kmo4OIzguP+dNZHmLncniBKjkckVU1ZIXx/15AhRmvK1P7yWdFdm48JGYcrVO3wFt/7gu9jT0Wto/dQZShBEsUNCjTBEKvVCRhsKFA+weLuHTbdvwvPLn49uU7y/ZIPcs4WWrQWHvnhMR42dnZXiUs+fYYanFctb9kcFjlFxvPfMXqyYskLztZWJPplANiISG2Y2YFzJuITtzBLEUPmv0LT7uCGxRp2hBEEUOyTUCEOIBJTR9JLRhgJFXCiGrV3ru6KzLdU1SJxzXe+vQsDGbJqF/HowWDDctyZsJAug1+uLChyj4liZZqD12prtslTvrzVuTOaNx+xe+PxBbN13ytS1jGwnCIIoNKiZgDCMzC/LCHoNBeoIzqbDm2LqzqzMGu36NFq/li60zF5Txc/9aLmjJaYZwWl1GooU2i12WC/ej4FLc2K2KwLnkCexEF/23PUN9mHbO9ukRfiNCxsT6sC0UGa7tna34umDG+HnI9HrPH1wIzrOd6C9p11aF8f9YVsUkQ+bkbVRZyhBEMUECTUiK2ilopTC/IaZDfjqvq/i8LnDMfeLrDmyAgcQKgFnATBLZurgWo60oOm2pqhAqttVB9+gvkDhnMM7NCq8TyZwXKUuaXen1sxWdfelEaG898xe1F5Xiy2HvxcVaQp+PiKs61PgITtGLtQDkPuwKSjdnuqZo+rfJYIgiGKAuj6JrCDr8nOXu6PpzdbuVngOeJK+htPqxHBwOKmZlVrwEAPnNjCLPyPRNaVLEYCpx88C4zFw+lsJ262M4eH/8yJ+ffaHMZEmG7OBMQZ/SN7koH49RBjt4LUwC0KhkOGRp5wjZqi90/lzeEQAACAASURBVG7FljXzYK/sFFpvmO32JAsPgiDyHVnXJwk1IisY+WA1aimh4Cp1oX+kP+EDPBWxJyOTKVAg/FiGA8MmPeEYAme+mzBPEwDGzXoOzDaUsL2ypBJl9jLN51lp1hgYHUh4bmV+eiLMPWcMFX0/wPnQW3Be3wZu88JVWokro1di6hGV3xlZdE8kNMnCgyCIQoCEGpFz9KIaRqM1CoroODd4DpWllVFx4bA60t4RmmmhlgzucjceveElfGPnuwiq/o5tFR1wTN4hXK/ie2dWFGcaJWVppBZOy/tN5OtnJJpLEASRa2RCjWrUiKyh14ygVfAuon+0P1qEr6698gV9sFvsKLOVRaNCvoBPWJ9lVIDlm0gDAF/Ah2e6/hyOmZXRlCEAlE7aJ12v0g1ptkFAC7vFrplO1UMp/hd5oonQao4QdXuShQdBEIUM2XMQeYPMBDUZ/CE/LvsvR2/LiugZC4u1NJe1ZQXviBccHJYSLxzu3bBVdAAI21vIULohO853YCQ4It3PKAwMzy9/HhZm7K3EzkqxbvY6oRWIUeGkRGNFxsi+gC/BCoQsPAiCKGQookYkTToLtFu7W7Hl7S0J/mipRGvUUwS0UMRaLoNmydWoXYVZ/CidtA+BgVpwvwusJFGsVZZUomFmAzYd3qTZeWmGqvKq6GueEKEL2bF4Yj0+Gjlm6HfESERVib7FzwGtKKnAUGAoKsjVXaxk4UEQRCFDQo1IivgCbfUHI6A/SFvrXGpSSamZIdepTc/icAOEmWL9eJRI2siFejjcu2NGWDmsDjTd1gQA2HnqtRRXe/WcithpmNmAox9+ite6fwLYvNEOzre6F2HLmv8aHdCuhV461lXqgmexJ/q7pE6l1+2qi6bBFZRRUiLTZOr6JAiiUCChRiSFbMbilre3YCQ4IhRwSueg0rGneF8p/49V1s1eFxUNqXas2io6EBiohd/5R9jHvw3GOCzMEp0MUberDpxLbDM4DIcVRX5lbUeqccUbu/4Awga8RoSanl+bd8SLLW9vidlXQa8OLRWzZoIgiFxCXZ9EUszbPs/U/mNdjMkos5XBF/ChqrwKK6asSDklGQqUYeTjLyZE1Ix4qAFXBZiWua2r1IUD9x+I3t7T0Yut+06hV2K0ywB80GJOJGl1pYqsNaizkyCIQkfW9UnNBIRp1PMajVIoIi0dg9LNMBQYAgdH32BfWurGmHUIjsk7Y0QaEB4gbySNfG7wXHTWqgx1Y8aejl407T4eI9JsFR0ov6EF427yoHzWcxj32eeEsz711iFDSWmqETUXUB0aQRDFAKU+CdPEf0iq0RpTVAgwsLRPNsgm4Vq75NdfUVJhav+t+07FGO7aKjpionmK6S6H9piqePQaC+KFXHxzAdWhEQRRLFBEjTCNVrTDs9gDu8WexdUgbZYeABBCYUT+MsXl0cvRqJfTKp61qd4eP1e0dNK+hGieGiUa1trdirpdddJImyhCpkZkraFEArvWd6FtbRuJNIIgigISaoRpZP5T7nI3gPDA8GzhLndj0+2bsna9YieEEJ5969nwcHjJdIcSa0n05/jB6VoebgpKZK1vsC+a9m1+qzlBrJVaS4XHU0qTIIixBAk1wjRa9UAtR1oSvNAywbrZ63B8/fFo5EQRiYQxKksqpff5gj7NtKPaBmND/Ww47dbobe536V7bwizCjmHPAQ/qdtVh0+FNaH6rOcFuA0CMQS5BEMRYgGrUCNPI6oEA+QSAdLPz1E785oPfgDGG/pF+zTQZkQhL0TiutbsVDTMborYbW/edwlmvD2WDX0SwdCf8XD71QKuxRKupwsIsVHdGEMSYg+w5iLShZangLndL520SxqgsqQRjLC+eQ8X2QjSdAtD2g3OXu5MeCC+y5iAIgigGyJ6DyDhaTQaNCxvRP5KYyhrzGPie5LA60HJHCw5+5WBeiDQg/ForEyXUtWaeAx50nO+QpqIVn7ZkG05E1hx6jQkEQRCFDAk1QhejH4RaQ649Bzwpp9tSQdbBaIZMeKxxHhkKr4G6UzJfqCqvEk6nAIAdp3ZgxZQV0jrGhpkNKLOVJX1t9RcCkVgUNSYQBEEUKhkVaoyxDxljxxljnYyxo5Ft1zLGfssYOx35f3xkO2OM/ZAxdoYx1sUYW6g6z/rI/qcZY+szuWYiFjMfhHqdeLkyvW25owUbl21MqY7NYXVkxF+NWYzNGe0b7EPTm99J+/X1EAqqkB0f/GEF+q7I05ftPe1oXtYMd7kbDAyVJZVw2BxoOtAknMtpBvUXAtkoMy2vP4IgiEIiGxG1uznnC1R5Vw+A33HOZwH4XeQ2APw5gFmRf18D8CIQFnYANgK4DcBiABsVcUdkHjMfhPlaN6TMeYwXDkbTbwwsOiszV3AOhKyXsnvNYAkGR2M90TgHRr23wD9Qi5BGh6d6wsGWO8LzX70j3qjYT5Z4aw69GZ8EQRCFTi66PlcBuCvy83YA/wrgW5HtL/Nwd8NhxpiLMeaO7PtbzvmnAMAY+y2AzwH4ZXaXPTbR+yCMLybPt8kEFnb1u0j8YO7W7lY0HWjSjZRx8LSMd0oFxvRTpGnHMpoQ7WMMsI07iZGPgZEL9XBM3iGOCAZcmOFpxWSXE2zq94Qp0njsrBTlJU7h7w/ngCU4Hl/4zFdjXkPZBAOtNDxBEEQhkemIGgfQxhg7xhj7WmTb9ZzzPgCI/H9dZHs1gI9Ux/ZEtsm2J8AY+xpj7Chj7OiFCxfS+DDGLrIPvKryKmFa9MrolaxPJtBCK93aMLPB9MikXJKuEj+lLi4UtIOHzL8FKKa2gYFa+C8tSRCQSvSv7IYWfBx6C97R89J1hEZd0f+H+9bgz677WkKKmofsGD67DgOnv4VX35yEPR290ftoxidBEMVOpiNqyznnZxlj1wH4LWPspMa+oo8hrrE9cSPnPwHwEyBsz2F2sUQijQsb0fxWc0xERPkgFKVFAzyASnslyuxlKaW40oWFWbDp8Ca097QLZ0CmUitViIRGXRh8/6p1hq2iIzzEnYn+XBhEf2pqU9uRj1ejgs3C+Cn/En69+VVByUq8cLh3A6ESwDqaeJ5gWcxaAOClNmBS1X2ovG4f+kfPI+R3YeRCPQIDtQAAnz+Iv9vRia37TmFD/WysrqUZnwRBFDdZ81FjjDUDuALgqwDu4pz3RVKb/8o5n80Y+38jP/8ysv8phNOed0X2/78i22P2k0E+aulD5JXVMLMBNdtrhGlDBoYtd2zBtne2aYq1XA1AtzEbxpWMy6sUrR6cpx5R4yE7hvvWREWPQvwgdWVfv/cW2F3HErarz+G0W7FlzTysrq2W+ujJ1h4KODF4emPSj0d9bYIgiEJH5qOWsYgaY6wcgIVzfjnycx2A5wC8AWA9gJbI/3sjh7wB4L8yxl5FuHGgPyLm9gH4tqqBoA5AU6bWTSQSX9ulIKsPqiipSIjCiagqr8K5wXNZF2sBHigokQYAVlaKYMgPZjHXOct5OCrG4yJTagIDtRhGZKC63Ruzb9A3LbqdBcfjlmu+goOXpwHgsDKGL91SHRVKZgv4mVU8S9QoPn8QW/edEgq1PR290WkJk13OSPSNBB1BEIVHJlOf1wP4XxHvLBuA/8k5/w1j7N8B7GSMPQLgTwDui+z/zwA+D+AMgCEAfwEAnPNPGWPPA/j3yH7PKY0FRG4RpUUBY+lEJX2q5WCfr5TZysA5lw4tFyJL4kewW+zwh/wxtznn0bmpIYwAsCIUcIJZfeBBJ5hlFMwSvHqJuMiVLIImIjBQKxVx6u1H7FYEefiaQc7x+rFeLJp2LVbXVkuFuyyFCgDlN7RIBaQRznoTX4M9Hb1o2n0cPn94nb1eH5p2HwcAEmsEQRQcNEKKSAklLWq0Ho2BxaRPtcZO5SsOqwMOmyOtUTklHds/0o+q8ioM+YeEglddY2ar6IiJggWu3ATbuJMJUTEjKFKq2uXEpcERDPkTI3dWxhAUvF9Uu5w45LkHrd2t8Pzb04AghVriOhazXY1RQRn/eEcu1ON6yzIc8twTs9/ylv3oFQi48WV2lJXYKMpGEERekvXUJzE2UNKiRgSXMh9STePCRjx96OmYaFK+MxwcNmQ3YYYAD8Bpc+LA/QcAADXba4T7KR2XgDgKNvKx+Px2CwMY4A+Kv5gpIu2Q556EiBQQrgdT31ajRLUaZjbg6Ief4rXunwC2q2LK7luEtfNX4Lfn/jv6R88nRBaZxY/SSfs0hVp8HZ3SqFA3bZp0PfFcGvLj0lD4eIqyEQRRKNAIKSIt6NUnySwTUh0nlO+YCVj3DfZh0+FNqNtVp1G3x2Cr6DC9jnWLP4PyEu3vZYrAWV1bjS1r5qHa5QRDWMApt0VwhKNYezp6sfGeh7DplldReW4bht734HrLMmxZMw/zx9+Di+9tkD8qu3Z0snTSvpimBiAs8A59+krCvpMl64xHqXEjCILIZyiiRqQFeX3S1UHcAFC3qy6he3RgdCCbS80OHGHnfssomG3I8GF6xrqMcTgm7wC//g2MfLzScGrz9WO90oiYgoUxTPe0wsKAUERRuZz2mBRhfKRNIT5CFR+lWt6yHz5/EOV+F1iJwNBWZfkhSrHKhFzfYF/C79SG+gXSdcYji74RBEHkCxRRI9KCzHi05Y6WaLpTNjPUjIt8ZUllWtedKSpLrkPoT09i5OMvpn2iAGOAxeaDw73bcHTNiGhRxFFItV6vz48Nr72LPR29MZE22TVkESpFEI1cqAcPxRoi85AdIxfqo7dDnCdcg2uMq4r/nbJXdiZEBF1OsQmz0egbQRBEriChRqSF+Fma7nI3mpc1R209ZDNDPQc8GPIPwcqsuteoLKnESHAkI+tPO2wE9999AddblsF/aYms6TG1S1j8KL1uX/pPHIc/xKMCbHVtNQ557pE2sMoiVIogCgzUYrhvTcJEAnVkcLLLibtvmhRzvEjgiVDm0Crr/KClAYc896B55Rw47bG/Y067FRvqZ+uekyAIIpdQ6pNIGzK/NUC7hs3odIBCmiLQP9qPXR99F3AD08rdWDFlXXQ6AmNMc7SVGSz2fs1C/3QRL8Amu5zCzkpFkMX7mN1906Ro+lVpglAaHAKqBgen3Yq7b5qEX779Ucx5Fa+3suvbAJtXM9Uu+l1TUrHkrUYQRKFB9hxExlBPNEinOClEHFZHNMKozEgVdY6anUDgLnfj0RteigoQSIa3y53MjFEdEVu/fPsjBDkHC2usmDSpMikAADa89i78qjvtFoZ1iz+DN09eiBFKAKSCTgQD8EFL+MuArNNY1F1MEASR78jsOSj1SWSE+IHtqYo0u8UOpuUYG4eN5VewWEnJAeHI46obV8HCLNEB6co/GaL7HVYHll/7UFTouMrswmfIbmF4cMnUaM3W+DI7nHbjf/p2C8P0CU78j8N/itaxcR4WaWV2S0xn6OraajS/cSJGpAHh9Omv3+2LpiM31M/G1n2n8PUdnQCA769bgEOee/DmyQua0UF1TRkNZCcIYiyQX59mRMESPw90yD8kjBhZmMW0aGNg4JybGjUV4AE4rU5z0wMyjNKhqI4CiaJnoqgai0TKQiELLCwE9zg3ll/7EF59cxJ8/vBjVDzC4glyjl8c/hMmu5x4cMlU/PrdPvgEhrYyxjlseKtbPAxkyB/CD9YtiEkhen3idSjbZZMDjv7xU2E6VSG+pkxd/0gD2QmCKFYo9UmkjFYqLx5lYHu2TG7XzV6na3kBhCNwyrimfECWAuUcGDzZgg9aGqQO/Nkmfjj6dE+rdN8PWxqw4Nk2oZjTSs9aGcM/fHk+1ZQRBFG0UOqTyBiijk4ZVeVVWTW5be9ph6tUbu2gMK5kXBZWkx4mVp0AkD8eYPG2HOPLxN2Z48vs2NPRK424yUSa0241JNL2dPRiect+zPC0Rg14CYIgCh0SakTK6E0lUFDXD2XL5LZvsA/10+sTapniSefcznQgayhgDBiueB0zPK2wmOk6yDBq0bjxi3Ngt8auzW5laKhx4xs73zV9bnW0ToaSTu31+sBxNZ1KYo0giEKHhBqRMjLDWlepS+qrZsbkNlX2ntmLVTeu0jTLtTDxn0I+VgZwyxA4IByQbrcyU40C6UJd5L+6thpb186PMZxdd+tn8PqxXuGatah2OQ2lO7fuO5XQhEAjogiCKAaomYBImcaFjQk1ag6rA57FHmlht+iYTDEcHEZ7TzuabmuC54BHuI+swSGdQSuz1htGsTKGEOdRy4ut+04Ja9dsFR3hmZn2qwPTjY6g0kJkHBs/RkoZISWjvMSKEEfCIHijhrSyNHC+pIcJgiCShYQakTJGu+/iO0NX3bgK7T3tUuPSdHJu8FzUHkOEq9SV8fRnukQaD4br+9TCa/I4d+Q5r45aXqixVXTA4d4dHWzOSrxwuHdjGGEzWaWQX5mz6XLawRjgHfJjssuJwZGAsLbMypih1KSWYHLardj8n8L+a8ka0uoZ8BIEQRQq1PVJZAVRZ6hiAiuLcqWTypJKzckGDMyU/Ueu4CErhvvWAkCM8ArfZ4ez/34EBhYkWHWU39ACi2AYemjUhcH3w89/tcuJQ557hNd98J/+Nw69n2jRsfyGa/GLry7VXbesQzVd3Zzxlh9AYjcqQRBEPkNdn0ROkc363PbONrjL3cJjKksqhU0AZoxvgbD1xlBgSHOfQhBpIb8Lw31rERioDUfSLLFijFn8GCr/FfoFfmrMLo4WqrfLol57OnqFIg0ADndfEu4f3325oX62cNZmuiw31APj4w14CYIgChlKfRJZQdYZem7wHLbcsUUYbWu6rQlAWOT1DfZFzXIrSyulaUoLs2Dx9Ytx5OMjCPEQLMwCu8WeV8a3ScGBwTNXI49awktdbaekR6Wn9V+1LhGlCZVIlQx1c8Cejl48+6sTMdG8Xq8Pf7ejE+PL7PjSLdUJI6TSKaTUdXHKrNGv7+ikuZ4EQRQ0JNSIrCAboq34qgHaNW5qg1yZSHNYHVh14yrsPbM32hwQ4qH8EGkcMBkIjCHkj/WC434XmCCVqRZe8XVpCfuG7Bi5UB+9vaF+dkId4aWeP4PPP0e6Lmuk8E6UelRzaciP14/1ZiXKJZt8AIDEGkEQBQcJNSIryDpDFV+1hpkN0g7RliMtulMMLMyC5mXNpsx3tUh7h2YK57Jb7LAOfhGDqm0jF+qFNWpq4SVKjwKRuaFxXZ+MAe9e2o9fn/hh9PnrG+wDr3wVtqE10u7Qr9z2GQBie4x4fP4gmt84EW0YqIxrWEgm6qVEztRROi2rDhJqBEEUGiTUiKxgJGoWH81R7jfSjck5R8PMhqQbE8qsFah0lOPc4DkwxhCCuXmk7nI3Ph76OOXh88K12crwzTsfjIkSBQZqMQxo2m3I0qMAog0ECpwDr3X/BMweK3KZxY/SSfukQu3Nkxewp6PXsA2G1+ePdo+qu0jV8z6NpkdlkTOZYCSrDoIgChESakTW0IqaxXeF9g32ofmtZsPnVgx0pd2bGqlHHrIjdGkV2v7u7wEANdtrDF8XuBoZTKV7VavrVJni4LBboiLE5bQD/sXwvi/3QTOSHo3Bpt9wEI8ijlxldulQeKP4/EH84vCfos+CXspSFjlTLEbiIasOgiAKEer6JPICra5QPdQpVJnYkfV0cg4M963BJ+eu1mGZmprALdGJC7LuVT3c5W50re+SHl9hn4Sm3cdjhNBIIIQvzHcndFKqGblQDx6KnbsZnx6NuU8i4MosE6PdlFZBPtjnD4JzaK7FKOrXyVbRAcvUzXj63c+h5mcrMOvbm2JmeMoiZEHOhR2mRs1zCYIg8gkSakReoNUVqkX8aCoZ8hIxhsBAbUy0pXFhY4LAEcFDdvjO3gcAqNtVp2vcy0NWWGCN22bHpZ4/w56OXjQubEywI3FYHRg5Xy+MHP3y7Y/wpVuqoyIqnsBALYb71iA06gLnYc+04T55vZlM2PFP/xyHPPfgg5YGhCS+i/0+f4I9xn9eMjXmtmxYuwilEcJS4gUYwG2XUOrejY9Db0VneMoiZIo1B1l1EARRDJDhLZEXyISOu9yNIf+Q0Ky2sqQSB79yMGbbvO3zTF2XcyD4/taED/Jbf/BdDJX/KpL2C/v286ATAAOzDkVrwiwAKqfuTWxg4EBIsP/4shKUXrcP3tHzMXVlijmrvbIzoU7vb34iX7/a1FVmKmsG0Zip4EAtPmgJC2HZNcaX2VFWYtOsLdPrDFWjZ9BbHbkGmdwSBFEsyAxvqUaNyAtksz+H/EPSjk8mSMPJRkHJasC434Uv3ZLov3XBOwcMc3RtcMtvaBF2mYb8roSCfQD4ZACYPLwIV+LEjs8fxLO/OoGyEifOehsx2eXEo/Wz0TCzGo+xf5YOM1d3M95906SYGi8zVCtjogZqhRG35S37saF+tlAc2a0MV4YDuDTkh62iA94J+/DUu1688N51aFryeDTaqTzH39j5rvDxhIfJM/j8QV2D3rNeX/R8yY6dIgiCKAQo9UnkBQ0zG9C8rBmu0tg6qf7RfulUgf6RxCibZ7EHdktiio2DI14bKPVab568AOBqxEeJGKn7D0S1WbaKDlgMOP6rmexySmurLg350ev1geNqIf2ejl6pSFM46/VhT0cvXj/Wm7RIO+S5B80r50jrzNSF/fFpxfISG/whHpOuZAzo95+H59+exrP7X4meZ3VtNf7hy/OFNWRb1tREzy2rl1O2K2nP1bXV0bTsIc89JNIIgig6KKJG5AWKNYeZweiion+1DUh8KpUxRMWaOu14FmHhJOoiVIRPhdOGK8MB+EPhLYookXaSCoQGA6I+X0ZSlEq0rFoycFyh0mk35GMmQymyV0eo1NdTp0OfeWc8ttz9LRzyXK0JnO5pBSDxbbP48Vr3TzB//FURpRcJW11bjdZuX0KEVRHW1BhAEMRYgiJqRM5RrDn0ivHVqDs942mY2YC2tW3CLkrGwiJq8H1PNMWnRGe0fLYuDfkBFrbFYADKrm8z7Pgf3Y6wOLn7pkmGOyR7vT7hnMz4x5SsR5jLaY+JQikRKkV/xkfJuO0Smt9qRmt3WJzt6eiN7iu18bB5sXXfqZhNepEwJcIafg1ZZM7pGvDLC2NS1QRBEMUOCTUi5xiZJlBZUgl3uRsMzHCnp0z4qQWFEuUC9H22/EGO8lJbuLBe4jmm2H3IOit7vT68fqw3pluz2uUM+6KJ1hr5f8saeZOE4uxvFgageaV4PJRyPlGUTG2bsnXfqWjUUStdmYyQbJjZgEdveAmBM9/F4JmwsA5yjteP9UYtOowiGhRPEARRCJBQI3KOngWHMqC9bW0butZ3oW1tm65IUyI+IhRBwQA8uGRqNDqjF7kCrkauZF5r3O+SijQFnz+IN09eiIkoNa+cI8yiKlG41bVhYSdissuJu2+apHnNeOIfezzKcyGLkp290pfQAarl25as2azWOCijqGsP4+v/CKLQaO1uRd2uOtRsr0HdrjrN9zqiOCChRuQcLYNZo9GzeKRGuTwsKKyMgePqCCQgnI5TitllcIStO/qHBxPv0zCTjSc+wrS6tlraCKDsKxKSSr2W0hBhhGqXE99ftwCbVsujdMpzYQmOF97P/a6EujmZb5vdtyjpmjJZJM5ohG5PRy++sfPdlMVetqEIICFCXSbCwaMTXEisFTck1IicIzN6bbmjRTd6Jvt2KY3SMcDuWxTtpIyPrii1Uz9Yt0AYXbNVdMBX+SqGggPRbZwDoYBTM+UZjyjCpBUxU9YmM3I1k1o02h25urYaW+7+VsJroyVIAwO1GP2gCfY//QOG3vfgesuylHzNZJE4IxE6JZIm65rN19mfFAEkZKQywYUoXEioETlHXTguqkGTiTHRt8unDj6FO169QzpKigXGG4quyKJropotxgCESg2LNFnXolbEDLjq8SbqlDSaWtSKFopQvzZGphtYGcPW++aj45m6tFhm6D0nWuh1wubr7M90pHuJ4iTZCS5EYUP2HEReIBvYrjWsXfTtMsADUosPh9UBb2+d8D5RdGV1bbi7cIanNSr79IxYZVgYEOKIOuqLxIuWbUW8q3/8wHKREW08ydpaKK+NkckHIc7T2pGZiqmtVsRM3USSb6Sa7iWKl6ryKmGTlKn5xETBQUKNSAnF/0w98shsPZkWWqF+M98i3eVuNC5sxLd7nOhF4geeVnRlssrHjPtdYILRRrKOR4UQvyqUtESGIg7j0YqyqI9RC5q7b5qEN09eSJtrvxExmIkolew50UP9usXDgbQKynQiW/dkl1MzqkrIyfT7VLYQTXDRsioiigMSakTSaEW70vUmqBXql327jIeBoW1tGwDAX584b1Iv0qQWKCMX6uFw745JfxptIlCnr8x+2BqJsiQraIwSb4gbnoB6lXwzot1QPxtf39EpTIKbTQFnE9kM07tvmqQZVSXEaL1PASgoAac29C6UNROpQ0PZiaSRDVK3MAu+ffu30/LmoTWsXTYfVLSvItQA7VovGUr3YJBz6eByo39JTrtVOEgckAs4WdpRSaXmIspSCNGdp/YcT5h/WgiD20XPrWyihTICjBAjew9xlbowHBhOiE4l02VOEOlANpSdhBqRNDXba6RF++l6w4v/Nhx/bnVKo6KkAkOBxCHurlIXPIs9Ka9FXasmwmm3YNgf0tzHypiwC3F8mR3D/pBQwIlq1JT7v3RLNV4/1is9Lt0UgjiLpxDXLEL2+8eAsAkzIUTrfUpE/Bc7gsgWMqFGqU8iabRSj0odWariyEyov8xehs/N+Bx+88Fv0D96dWC7d8SblpSsVs0TAPj8Ic3j4yNpai4NJY6j0qtBU6IsWrVr6USvoSFfyXRKWEa6BaJW7ZoZiqVeyyhGSyQUqIOSyDcookYkjSjapYaBoWt9V1av77A64LA5hJ2fqX5TFkW11IhSovzyQoQ4101faVHtcko/7LWifAxIawRJK/1KqbdYZBHQVCKd6Tin3AIewwAAIABJREFUXoS6GJE95lJracwXOgWKqBG5QhZRIx81ImkUjy0LE/8aZbplXNYRKrPn6BvsS8nBW/FWG1+WOJczfni5pcQbvn3NOzF+YkbGVKlhgKbxqVY0Jd1mqWQbYZxMeKFpGR4bZSwapsp8GptuaxIabVMHJZFvUOqTSAnlW3guWsaTSVGYTYGK0lcdz9RFtysRJqERrsWPkuv2YYanNiGyZSSyFt9VCSSmNY1YZqQjFbqnoxcWSX1dvhrH5pJMidpU07hj1TBV5tMIUAclkf+QUCNSJlct42ZrTwBztXNGarL+bkcnAA0jXJs3JrKlHLu6tlrTQLZaox6u1+vD8pb9CcLvbCTyJiIVgaA1iinfLDnyhXTVk6Ub2d8MYww122sKXqyYrb/TEnAEkS9Q6pNICw0zG9C2tg1d67t053OmC9GMUCMYjR7I0lff2PkuntpzPCq8ALnhrXp7fOpLNh7pB+sW4JDnHk2vL3VKU5lP+kFLg+680GSQjWKyMpb3Nhe5IpXRV5lE9jcT4qGCH/JNA8uJYoWEGlGwKLUnDMzUcUZr52RRqCDn+MXhP8WIl5EL9eCh2No1kRFuvEGtVs2RXj2bqOYpEwJB9jyke1xUMZGOerJMEF+vJaovLdSatbFYf1csyOY5E2Eo9UkUNA0zG7Dl7S3C7i0ZRmvn9EYQqQkM1GIY4Vo1i90LBMdj+OO6hOHlFsaiUTBAu+bISD1bvIhKZTamjHxN4+U7ubIF0UOd7qvZXiPcpxBr1sZq/V2hk40JN4WObkSNhfnPjLFnIrenMsYWG70AY8zKGOtgjP06cnsGY+xtxthpxtgOxlhJZHtp5PaZyP3TVedoimw/xRjTn9VDjCkGRgdM7W/0j99sh2ZgoBaD73tw+WQLhrs9YFcWJuwT5NxUF6aS1jST0lSnQpVu01QQPQ9KN+rylv0pdZTu6ejF8pb9mOFpTflchHlk0eVCHPJdTI9lLEGRUH2MpD7/EcBSAF+J3L4M4L+ZuEYjgPdUt78D4Puc81kALgF4JLL9EQCXOOc3Avh+ZD8wxv4PAPcDmAPgcwD+kTFm/NOTKHrMvBG7y92G91XSV1YmTq1qJVz9QY5xDpvw2GRsGnJZ86RO4wGx3aip2H8oTQpa9iNEZhHVrBWqRUUxPZaxBEVC9TEi1G7jnP8NgGEA4JxfAlBi5OSMsSkAGgD8NHKbAbgHwK7ILtsBrI78vCpyG5H7743svwrAq5zzEc75BwDOADAc0SOKH9EbtI3ZYLfE1owl86a9urYa//Dl+UKR9OCSqZoF/5eG/MJOScB8F2aua57UkT2ZZYhZMuE1RphD5jFWiCmnYnosYwmKhOpjpEbNH4lgcQBgjE0CoD0r5yo/APD3AK6J3J4AwMs5D0Ru9wBQPmmqAXwEAJzzAGOsP7J/NYDDqnOqj4mBMfY1AF8DgKlTpxpcIlHoyOxBRNuSedPWq/vSstmQoVXfJRs9lA81T+n0ByMD3fygmCwqiumxjBUaFzbmxIezkDAi1H4I4H8BuI4xthnAWgBP6x3EGPsCgPOc82OMsbuUzYJduc59WsfEbuT8JwB+AoRHSOmtkSgeRG/QqXYOGZ3VaMR0Vo1WyjLf52km21gg8reiJgWCIHLlw1lI6Ao1zvkvGGPHANyLsGhazTl/T+cwAFgOYCVj7PMAHAAqEI6wuRhjtkhUbQqAs5H9ewB8BkAPY8wGoBLAp6rtCupjCEJIqp1EZgST0WkDRmZvZnPIejKIRKlerZzstfjC4sfw6puTTJ2rEEn3cPZCZ6wNhSf0oUioNka6Pl/hnJ/knP83zvmPOOfvMcZe0TuOc97EOZ/COZ+OcDPAfs75gwDeRDgqBwDrAeyN/PxG5DYi9+/n4YnxbwC4P9IVOgPALABHTDxGYgySaieR2foppYZLCyNdmPmeDkymVk72Whz69JW89BpLJ9QwEQuZ0hKEeYykPueob0Tq1W5J4ZrfAvAqY2wTgA4AP4ts/xmAVxhjZxCOpN0PAJzzE4yxnQD+A0AAwN9wzo3lmIgxS6qdRMkKJpfTDq/PL9xuhEJIB5qtldN6LTJZd5cPkax8j5BmG60vUBRRIQgx0ohaxLvsMoAaxtgAY+xy5PZ5XI2CGYJz/q+c8y9Efu7mnC/mnN/IOb+Pcz4S2T4cuX1j5P5u1fGbOec3cM5nc87/v6QeKVGUyBytU+0kkgkjPcHUvHIO7JbYskq7haF55RzJEbFk24YjGz5muejqypdIVr5FSHPlAK9cVzabl6wYCEKOVKhxzrdwzq8BsJVzXsE5vybybwLnvCmLayQIIVpplFQ9lZIVTKtrq7H1vvkx6byt9803HD3Jpg1HtsRMLvytjKSusyFSkxX8mSBXaUf1dWWQFQNByGFc4vMUsxNj1QCmQZUq5Zy3Z3BdKbNo0SJ+9OjRXC+DyCCyb+jucjfa1ralXLScD6mzTCKzFal2OXHIc09aH3+2C8hneFqFreEM4VrB+GYRICzE0y2Ks3UdI+j9vWT7ugoOqwPNy5oBUOcfMbZhjB3jnC+K365bo8YYa0G4Xuw/ACjvNhxAXgs1ovjRq0NTdxIpQqHpQJPhD4F88C2LJ53iSSstl26bkGx3denV+mWrdiwTs1eTJRMO8EYEuNb53eXuaGS1GOc9FvuXPSI7GGkm+E8AZiu1ZASRL1SVVwm/qcenUQpt6K/szT3d4klLzBR6EbyejUg2a8eSFfzp/pA3+vdiFKN/V7LrqiN5dbvqUm4yyDfbj3z3RCQKByMjpLoBGGtZI4gsYrT2KVdDf5OpgdKqG0v3yCWtOrx8K4I3i16tXz7VjonIRP1gumsFjf5dGbmuLOrWN9hnqOkh17YfoiYNGpFGpAsjEbUhAJ2Msd8BiEbVOOePZWxVBGEAo47WuRj6m+y3aa0393SLJ620nMy8N1+EjJpkRm4lY9ybTTIR0Uy3A7zRvysj15VF3QBjEfBc2n7IIove0CoAtQn7F8qXHSJ/MCLU3oj8I4i8w0jtU7pTPkZI9oNWS4xlwmNNJmbyXcgoJCuI01U7lqkapExFNGV/L8mkDc38Xen9nYrmParRE125+DKmIBOJzuvb4B9IFGr5+GUnG+RbarqQMDJCajtjzAlgKuecYrZEwZGLob/JftBqiTE98ZRO0ZBPRfBa6Alireck1WaRdNYgxa+zUmKcLPqQT/UDMNkaznT+Xamjbsl4rZkRjekWDLJ1cZsXTrs177/sZINCqxPON4x0fX4RwAsASgDMYIwtAPAc53xlphdHEOkgF0N/k41+aYkxLfGUicLlfOx6jSebnavxpCs9KVqn3cpgtzD4Q1dNRpx2K+6+aRKWt+yPvv51i3vx67M/TOkDMNm0Ybr/rpSom8zOQysCblQ0ZkIwyNO2HBNu3oqR8/X45NycvP2ykw1oIkVq6PqoRQay3wPgXznntZFtxznn87KwvqQhHzUil6Tin5VMZEzPEy3T5MqGQOtxA8joc6Ln1WYU2WMYX2ZHWYkNZ70+uMrsGPYH4fOHYvYZd2MLmN2bcKwZb7Sa7TXgwkcCMLCsp6nixRRw1WtNaw1GImWZ8JITrVeNbO1jKRUo+x1jYOha35WDFeUnSfuoAQhwzvsZixmLo++SSxBjmFRSh8lEsnLZpZlLGwKtCOTXd3QKj0nXc5KumkHZerxDfnQ8U6d6fkOJO9kSRRpgrjZLq5Bf3UEJZCdNlWykzki9aiZq2fTStqLI0VhLBeaiTriYMGLP8XvG2AMArIyxWYyx/wfAWxleF0HklHSMF1pdW41DnnvwQUsDDnnuyahoyaXdRC5tCLRsODL9nKRrLqveOkXPr62iA+U3tEjPaeYDUGSfEU827GzUNMxsQNvaNnSt70Lb2ra0iZdMzZ1V1svAhPfHC8FcWQblilyMkSsmjAi1vwUwB2Frjv8JoB/A32VyUQSRS/JloLcZsj3MXU2uPddkgjgbz0mp7epb6Pgye1KjofTWGf882io64HDvhqXECybQBWY/ABtmNqB5WTPc5W6p0ACKY3B6pgWDUSFoNrIn8mkrJOJ/x9zlbt1UNnEVzdQnY8wK4FnO+QYAT2ZnSQSRWwrRlT+XXZqZsA1JB5l8TkQ1iMOq1KSZmj29dcY/v6WT9oFZEjtCgasjmcx+AKrThskU8xcKmW4sMtrUYLZLtRjSpNkeI1dMGGkm2M85z3w1cpqhZgIiWdJVJD5WSPfg8UKYj6jVxCCrm0vl+VCfb9xNHmEkDQgLtb7BPliYBSEeSkq4JVvMT4Qx0iRg5jnORAMEkZ+k0kzQwRh7A8BrAAaVjZzz3WlcH0HkDfkaIcpXUo1cqYWZq8yOK8OBqC1FPs1HVK9T9vX2rNdnOiIb/8G+/NqH0HakOua53LJmXvTaluB4cNsl4fWVD/QQD0Vvm42+5MLOppgwEjky8xxrjddq7W6l12UMYCSi9pJgM+ec/2VmlpQeKKJGJEu6I0T5Rj5FrETPtYhs2YzIMLNOmZATRWRFkRUesmO4bw0CEVf7+N89PTsIERR9ySyZtNqQRdQAinQWG7KImmYzQaRGrYtz/hdx//JapBFEKugN9E6GdHSRpoN8a5QQRZ9E5Ho+opF1Kg0AZrpNRd1/zOJH6aR90dvxHbSiwmw9iqERIF/J9EB4ra5cI52ihd6IQOikPjnnQcbYSgDfz9J6CCIvSKcrfy59xuLJt0YJowIs12lnrXUyICEyaXROqkxAxZvYxl8/Pr02/+X50XSniGJoBDBLtgxlM+26r5zDc8AjvF9LhOdDI0I+RfALFSP2HG8xxn7EGLuDMbZQ+ZfxlRFEkZBLn7F4cm2lEY8RAZYP8xHV61Q8zMbd5EHFrO/gR19DjC2ImYisTEBxv0t6fRFaIm0s+lVlIsoli0xlYyB8w8wGaeRUS4Tn2q8t3yL4hYqRZoJlkf+fU23jCI+VIghCBzPiKNPfPvOtUULUIWm3MpSX2NDv8+fNN3BlnX7nUTjcu6P2GNx2SRihkEVk41/fusUP4dfDP0yoURu5UB+9LRKqCUPcp16Hfv954dodNm0z23wm2ahYuqNcWpGpbLnuG7X+UJMNEalFriP4xTKmS1eocc7vzsZCCKJYMSqOspEi1Rq5lAty6f9mBmU9z7zzbfA4DzMjAqC1uxVbDn8P3tHz4BNcsIbq0eutxatvTsL9dz+GQ5++Etv1+XE1zkL8fIh+T8rYvbBX7UQIiXV03hFvQfpupZK20+qUrNtVl/DBrfcFSUv46QkoLbFgRkgk041rVkSm+4tiLiP4+ZD2TRdGuj4rAWwEsCKy6d8APMc578/w2lKCuj6JfMFoF2m2BqtTzUjyJDNcWq+z0+zrK/o9Kb1+D0rGH4bGYIGC6/w04x8WL3iG/EPoH9X/iHJYHfjC5Mfw6puTNP8+9V53meDS8ksDkBa/Oj0hqHcN5fi+wXPg/koMn6+XdhybJVvvaSIK0X8uFR+1nwP4PYAvR24/BOAlAGvStzyCKF6MRo2y9e0znY0SY41k0lxanZ2BgVrTr69opJRdR6QBkFo85CtG03aiyIndYoeN2RDgAc1rDAeH8foH/wSf/1sx2/3Oo3jmnW/jmS4vqsqrUFlaCe+IN+F45XWXeafp1Yilmp7VixrpReHij2d2Lxzu3RgGEBioTTlNmcsIfq7TvunEiFC7gXP+JdXtZxljnZlaEEEUI0bEUb7VjxGJGKkTio9YDri1OzvNvr7CkVI6Ik3h9l/ejqbbmvI+9dPa3SqMYAGJolgkhvwhP1ylLjhtzqhAkQnVkDXWPFiZpaqkuPsG+2BjNtgtdvhDV9PeRpo0tMSCLJfVZ0JIGKnF0zLg1fsSAaT2RTGXpQ3Zqh3MBkaEmo8xdjvn/CAAMMaWA8itqRFBFAhm0oz5Vj9GJKIXoRDVj42bWJlgtwGEOzuTeX3jf09E55bRP9qf93U6rd2teOrgU8L7bMyWUPslE2D9I/04cP+B6G1ZKswSHB9zWzRLNcADqLRXosxeZqowXUssnOsfFk6YYAFXwjYZZqJGohSpEXuYVL8o5iqCn0zzRb5iRKj93wC2R2rVAOASgP+SsRURRJFgtjmgUArrxzpaEQpRl9vw+Xo43bsB1Yc/D9lRNvhFPCmp/9ES+PG/J1ojpUQMB4fxxMEnoo8l39j2zjZpynJcyThp7Vc88ZET2Qf3Fz7zVbz6oVVX+A6MDuDgVw5qrj1eDK2YsgJ7z+wVioXH9+6BffzhmGgo58DogHHhbjRqJEuRVpRUCGv5FHuYQv6iWEyj0HSbCaI7MlYBAJzzgYyuKE1QMwGRa3JZSDvWyJcGiRmeVmFKy17RgRmfbTf0gWF2hFkyI6WAsGBYdeMqtPcYW1e2mLd9nvQ+pXj/jlfvENaMKagL9uOFk+jxqn9/rpn1HaHw1StClxXuy57jmp+tkETUxqPrkfboObWEhtHh7rJooqvUheHAcOzvTqTR5TrLsrT8HRWLRUY2SLqZgDH2bQDf5Zx7I7fHA/gG51wcmyaIPCXbbxj5Zi5brIgil1/f0Ymjf/wUm1bLP/QzgazO8DrLMrStFb9lxovMwZGAKe+phpkN6DjfgR2ndpha63BwOOaYfLAv0DOk5eC6Is1d7o6mt+KjSHvP7BV2VarTc63d4m5MvZSZrF6svaddKPC4TfwYlO1G7CWMRo1kKc7+kX58efaX8dofXsP/3965x8lV1vf/853Z2d3ZxOzkhkk2IKRiaBHIkojURFpAk9JwyY9LUBSotdr+2v4IYIMbRQgIJoIV4ddqf9RLsWJJuLgEU5tQoEqwqAkbgihUjQLZrJKQ7IYks7uzO8/vjzNn9syZ5zmXuZ7Z/bxfr7x255lznvOcMydzPvu9ZlUWMYnhst+/GDd+pDKP9/FUIqOeBOlMcJ4t0gBAKXUQwJ9Wb0mEVJ5q9+PTEabnIykdnbtRAbj/2VdrXgF99bL5SCbiAAo7GMhxt2vvNV3l9v50pmg7wFvg/2DPD7Tj4pcK6qKWVet1BDm2n0jbeulWLJ+3vOSq/O5eqqmWFFriLVjz9BrPXplhswxnG4La7fGg618+bzm2XroVu67elT93N6YA+inNU/DoLx/Nd7bIqiwe/eWjFfterHdnhPFCEKEWF5EW+4WIJAG0eGxPSOSoxxeG86Ft08gxH1HFJGAUUPM2XSs6O3DJwg405zIHY839EAEGMq9r/zAI2pQe0At8u62RV+mN9e9dj5gE+aq36DvSh1PuO6XsBt6lNAMvt3SC0+pVTnkGW/yse+86DI4MYmB4wPcPPJMYMo3rmq07LXeVLC9hOlYmm6nq9+J4KpFRT4L87/0WgCdE5KMi8ucAHgdwX3WXRUhlqccXRpiej6R0vCyU5bqZu3t6sXj9kzihazMWr3/S10LX3dOLh3f0IqHJHHQ/ALt7erVuUh06ge+0EpuYNWkWls9bjs8t+RwSsUSgY9mUY3Uu1YJdbukEpzUprHDSEeYPPC/hpbuP3Ja72ZNmF7hlK7F+G92xLnr7RTg6clS7fd/hvsD3vBeVPIdKUMofD1EgSAupO0RkF4D3wSqp+Fml1Jaqr4yQClKvmjosLlt9Vi+bj+s27NQG8ZfjZi6lpZdtIZtsyBy0/zC4sfsF3P/sq8ZjT21LoK25yTM5QicinLTGW3HW3LPMZSkQQ0xinkVhS+2PWWqvzVWnr0LX012hjmWTaiksa1GJ8gxh/sAzxYtlBhZ4xFDqM4g3796M9EixiPdbv1ccrjtbeelDS43zZDOpgibqQGlt7KJUIqOR4+WClOeAUuo/APyH7j0R+W+l1B9WdFWEVJgofWGQyrKiswPbXzmA+599tUCsleJmdgb2x0Qw6sqK1wX1O/ext1aZFKRZX8m+u6e3aK1Okok4br7gZM+yHb39aUw+qc9Y6Hb2pNna0hBOsshiSvMUz5gvoDSrc6kW7FITIxKxBJYdv6yoj+fa96zFuh+ty5egCNug3u8PPJ0wcicOLF7/pDGGctHbphV9zqYs3vbmds9ixWGFiOmzUAoY2rcs/7qc7gRRKpFR6h8PUSCQUPMh3J1PSB2I0hcGqTy3rTgFi942rawSHW4Lmluk2fT2p7F4/ZN5EegupQFYD7rW2Y8UuD/tPww+t/Flo0gDYHSPu9dnEoO2ZSmI2BkYGsDsSbO920uNpNDd04tE+85ApS6A8izYncd0YtMvNyE9qncL2yLUPvaU5ikYHh3WZrBe9PaLMDQ6lB83Nag3WaLOmnuW9jqeNfcsrTD6zDOfwbofrcOh4UP5efYadLAdQxnUUtqWaPP8vgorREyfkRpN5rsS2JQTQuBVd7CWNHK8XOA6asYJRJ5TSp1eofVUDNZRI4SEwVT3zkQyEUdrIoaDR/VZmk1Teqwq94l+pJqPwZozr8fyecuNtdYA7xp77vXZrY6cYjAuTRDAt8elTUxiuOwdlxktb3bz+OZ4DK2zH0FGDWlmsXDWDNMJgCANx03WJIFAQeVLb3g1HXefn53R6MRZE82rFpmp88HsSbMB+PdPbY23Avsvw77fnqx9X1dfb83TazwbwJvwaxzvRnvtsgmk+y4uEmqm+7KRaqQ1QpN2Ux214KlAhBAyjglrNUhnRo0iDbCaWh/9VRdWpL6NbVc8kX+AmeLmBDC6anWJByOHOjHYdzGywykoBWSHUxgdaQ4s0oCxcgwXvf2ivPiAiuXnG8w9tGXa9zxFGjBWl830MPQTaYDZmmQLEHdSgl+cnk6kAYVWFC9LlJcVJoglZnB0EC3HbCkqktI0pQeTTrwFrXM2FCVctLe0a+fys0aGDdzXJRhc+rbrkEgX6gRTCEE9Sh6Vg1+WbZSphOszXKEeQgiJIKZitXFNrFpQFICnXtpXMHb2STOLYtQEwIfOPM7o8rx+w07t/COHOgusH5NP6gr9hewsytrd04trNccK00/UTRiLRVDxY7vz/LY3WdSc4sU0R9+RPqNb2N7fz6IGWKVZzjr9VfzgueOgoLeE2gyODqIl3oLWeKtnPK2uVVUpyQc6t+RpU4N1+Wi0mK9GDn+phFC7sgJzEEJIXXE3OwfGWjfZAfxukokYAPGshea01NnlO3QiTddFobunF9dt1Ge06vDq++kVi2aLFVPdOVM8XBCCiBkbU9yUG3u9Xtu3xluxYOYCPPvbZ4veO2vuWYGO2T/Yj0QsgUy2ONYQKO5gYOK59JdxzuLz8dKL70f/9OLSLU4ODR/CuveuMwoKXWycLo7OL/nARNBM9UaM+YpKvFxYfF2fInKmiPxERA6LyLCIjIpIvt+nUuqn1V0iIYRUH6+6d6uXzUciVmyrGskqXLKwAx0Od6azI8Gk31uPGbNezL9n6qLgtroBY8kDQY15yUQcl5zwMa17Z/1712PrpVvH3JsubAuRyf07tG8ZElJanfMwxXZ17ikd9npN26daUlj7nrV45c1XtPs7Ozl4WZzSo2kopZBqSWlrnbXEx65JW1MbYh6P1B8f+C4+tTKNeHNxE3T3uXl1G/Bz9+bX45N8UC5Rq5E2ngliUfsHAB8A8CCARQCuAvD2ai6KEELqgcmasKKzA7c89mJRTFpmVOGpl/bhma5z0N3Ti09tvQ+xY8bcWtLcj9GWjdi8+2Qsn7cce/vTBUkGKpPC0L5l2NvfWXTMIF0LRAAoFLioFu2elrfGTEnMxNDry/C39wKfSz2JpWdcie8O3mN0q5ncv5MzZ+CzSzqNwfVemOLEdDjdU17Hsdfr585a8/Qa7f5hrD4jagTJpiSe/sDT+TFdIH5WZdESbzFmq9rr9LMC+sVMBV17tS1bLHlUOwL9qaOU+iWAuFJqVCn1DQBnV3dZhBBSO4J0IOg3JA7YVqgVnR2YduwTRW6tjBrKV7KfMevFgtZSseZ+tM5+pMDq5p7XiyYR3HX5AjzTdU5eYNrWmFtP/R7e+Plq7PvtyfnipQ88NRPnz7nGWA3f1PZs7YUn5+c1WeVMljPT9ibs43j1KXVairysT0GsPkHaJblFjyk+y0uk2fP4WQH9rGCmZAM31bZs+XVWIJUjiEXtqIg0A9gpIncA6AMwqbrLIoSQ2hC0A4HJ2uTM4jyUKXZhAmMP+pZjtmAwUyjkJJbB4Fu+i+6eDwU6npNMVhmLkeoscunMKLb+uAPPdI0F99ttdWyL1AfOvhJbf9xhDCY3WWqyKusbBB+G9pZ2bTFedwcCL4JYfYJYnqYkZhYUNp58Ul9JaXS2WxMotgLaY2ueXlNgGbQTB/qO9OXLlPhhn2O1y2c0asxXoxHEonZlbru/BXAEwLEALq7mogghJGyfzVIxCRp3YL3J2uQsXeBnwTEJORU/iDWPvFBwjrrj6TBZ3oKMb969GTduu7GgxEL3ni/iUyvT+PX65QWWOve5uLEtKpWysJhqfHrV/nT3cgTguyY/y5PKJrDv1bOx+sHn0ZvrPpHN6MVie3O7McbOKRDdVkAA2lIXtz17W0E/V5NIS8aTRedomtNdPqNR+19OJIIItRVKqUGl1CGl1C1KqesBnF/thRFCJi62lct+MPb2p7H6wefReevWigu3oELHK9nAxq9Wk0kUqEyqSBy6jxc39Isy1WVrT+qbsDu3X/ejdUV110bUCLqe7jI+uL3O0RYg6967DoAVI1bqw//Q8KFQ46a6XgCMrlHT+SiFglpyQ/2dyGTHRNLQvmVQ2cLr2xpvxZp3r8kLQ2DMHewnWk2u1Af/58FAiQODo4NF5xikoXyj1UKbqPh2JtB1HhCRHqVUcfRrhGBnAkIalyBdAuzSGX6lBJwuK50rz3Qsry4BXni5m3RB6Hb1/5FDnRAAd12+oCBxIZVM4PzTZuO7z/ehP13oNnVeA+dxpyRmYt+rZ2PIlaSQiAk+/L438MyBf8Vvj/w2kButSZowuXkyBoYGitx0Qc8xSFcCN36V5N3X+WjmaL6np257L5xzjQ63Y2jfsqLq/G7spJD18J4/AAAgAElEQVR480DZbsVT7isuzRIG3TkG6VTQCNX6JxKmzgRGoSYiHwRwBYAlAJ52vPUWAKNKqff5HLAVwA8AtMCKhXtIKXWziJwA4AEA0wA8B+BKpdSwiLQA+CaAhQDeAHC5Uuo3ubnWAPgogFEA1yiltvidMIUaIY2LV5slJ35iyh1/BhQLPN02ibhgUnMTBtKZkvqGerF592aseerzyMYP5rM+bVGQSiZwZHgEmVH/s5/alsg3b/cTgDZTZuxCy6xHAllpTPiJLtPDX0am4s1ffDLw9fQSfEDwGma6FkpeYjpMKzG/+y9IjNjm3ZvR9XSXdn9TwV4nps8jiAgL23aKVBeTUPNKJvghrMSBGQD+3jH+JoAgn+AQgHOUUodFJAFgm4h8D8D1AO5SSj0gIv8ES4B9JffzoFLq7SLyAQCfB3C5iPwBrPIgJwOYA+A/ReQdSinvvHVCSMMSJJAe8M+M9Io/s4WC/dO2uqXaEjg8OJK3XpmSC0pl+bzlyAws0ApIEQQSaQDQ1tyUX4/OzSWxDFpmbikQaqPt/16WSAMs91nXU5/Hp/81oRWyfaZkg/jBvBs7yPX0Kr2x9KGlgc/D6W7u7unF7d+/H+n2B/LZuU4X6fJ5y7WFjxNxARQK3J+m1ko2usK0uobwXlmnXn1YASuxouuMLq1oDpJIYSoVwlpo0cIo1JRSrwB4BcAfljKxskx1h3MvE7l/CsA5sCx1AHAfgLWwhNpFud8B4CEA/yAikht/QCk1BODXIvJLAGcA+O9S1kUIiT66h6UOU3yWTZj4M1s0LF7/ZFG9NLe4KwW3C/aShR146qV9BS7Z6wytonT09qdxfNdmpJIJjB6vF0fu1k+xMlpBOVHxg1ohCwAq065tOaVG2zDp99bn68fd/v0LsKLzBs/jmLIKg9YIcwoT23IaO+4xxFwlVJytj9zC3f5sdGNe90PQFkte53LjmTei85ix+nW2hc3dnF5HkJZJUauF5hemMFHxLc8hIhfDsm4dAyshWWDpsCkB9o0D2AGrQO4/AvgVgH6l8tGrewDYn0IHgNdgTT4iIgMApufGnT1AnPu4j/dxAB8HgOOOO85veYSQiOJ+WLZrXIJ+Fg0gWEkNN2GzKIOgKwHy8I7eohg7u1WVriiuKWaqP53BpOF2xDQtnpQjO1EAJGMzkFb7Sz4P3bxAYZbsYHZZUS9LlY1DYoOQpqPWWpr7kW56IF8IOCwmS1CqJYVkU1IrTGzr6mSDWHUKJq/Cx0EJ2mLJdC52QoJOrHb39OJzG1/G3/Zv9hQ0fuUzotT/MmiZnIlIkDpqdwC4QCn187CT59yTC0QkBeA7AH5ft1nupy6lSXmM6453L4B7AStGLex6CSHRwf2wLOWvbVP/Ti+BZxJ3MRGc0OX9YDRhcsHe8tiLBed09kkz8eBLjyLx1sLuBq2zH8EgYBRrQ/uKxVEczRjctyz/WgEY6D0XiVn6huBOZk+abSUlNE/B0ZGjBb0uVTaBIce8NraQVejEIFAgNBEbQqyp8JpKLFNyA2+TJcjkBixYn6FvaaXdfUHdimGtWpUWNNWqhRa2hluQMIWJShCh9rtSRJoTpVS/iPwXgDMBpESkKWdVmwtgb26zPbBqtO0RkSYA7QAOOMZtnPsQQiYIQZtFu/fZ/soB/NuPXsOoUoiL4JKF3vOY3K6jucSrIA/GG7tfKDjmqCFp6+DRTN7NalvZ2k98AmlVXBS37a1bccgg1EYOWeKo7a1bgaZ+zJo0Cwf3vA+ZQycXbDc00IlRNSaidBU/3Bl/XskPTmwrZW9/GiOHOgu2mXySPljeFM/mRymWIFuA60St20VaCfebSYAtnnYlFq9/0jH/Aqx9z9rA5xJ1QRMkDlBHNSzZ44UgQm27iGwA0A0rQQAAoJR6xGsnEZkJIJMTaUkA74PlQn0KwKWwMj+vBvBobpdNudf/nXv/SaWUEpFNAL4tIl+ElUxwIoAfBz9FQkhUqXZMSndPLx7e0ZsXSqNK4eEdvVj0tmnG47jdrjGN0PJ6MN7Y/QK+9eyr+dcmkaYjnRlFU3a/3o/Q1I8OjySLkUOdePNQJ3693noQntClr4Vli6imKT2egsXGlPzgxGml1Fkw2wwuV5VpR3dPb00sQXkBfqjQ4pdqPgZrzrwemYEFWHDL1oISKOVYq3RicvG0K/HAUzORzqQL5l938YLA5TCiLGiCxgHqKCVMYaIQpODtFABHASwFcEHuX5CCt7MBPCUiuwD8BMDjSqnvAvgkgOtzSQHTAXwtt/3XAEzPjV8PoAsAlFIvAtgI4GcA/gPA3zDjk5DGR1fU1l2dv1yCdh1ws6KzA890nYNfr1+OrEFo9fantWu93yHSSsFU9X7WpFlYvWy+lYFowPlQMz3gmqb0YNLvrUfrnA1Q2QSyI0lAeRdldRffndqWQCqZKCr8ayoKfPOSvwNcBWJVNoHB15f5fhaVwrm20UOdSL1xC2477T+w7Yon8kLUXacOsO6XazfsLKnIsrsDwdYfd5R0Pzoxfa5REDT2/zddMgngnTgRpPPHRMXXoqaU+kgpEyuldgEoso8rpXbDytp0jw8CuMww1+0Abi9lHYSQaFILF04lrA9epUJ01pZyg2PbjlwAJB/Uxiwtn2cd59PfeQFHhguvnfuhpnPhuq1o0nQUKpvApcfdgJvPudJzXUFdz/rtOnDtAz1o1iRI7EX1LEE6i62u7pnuXnQT1LrmZSWuxP1YStylF5Wyanf39Ob/n5QSB2jKto2CO7feBMn6nAvg/wJYDOs7aBuAVUqpPVVeGyFkHFMLF04l3ClnnzQT9z/7qlaAlSMsp7YlMJjJFj1wP/1HH0Ki/WRjzJIthPwesPbvaze9mLcUtczcUpRIILEMNv7qXmz9cYf2wViJB3l3Ty+yb56OIx6xbZUmTNB90HvO7/P2O2al3HutiVj+GKlkAmsvPLlkcVWJxAR7Hhu/OEATpcShTgSCxKh9A8C3MWbt+nBu7P3VWhQhZPxTi5iUcq0PdoybU6S5S2e8vm8ZrPKQFpOa40XWLjfJRBw3X2AF+utFUEeBC9JuUO/ezu+htqKzA3dueTkv1EwuKUn0ax/SN3a/UCBSS3mQ2w9xXaxeNV1bYSy2QQssA96izu+Ylbgf3fsPjXh3LvCiUlZt9zwjhjjAepT9GA8EEWozlVLfcLz+FxG5tloLIoRMDCrtwtFRrjvF/QAqch3mSmds3t2Zfwjd/r9OwScefB6jjir2MQGmtOor+QfpVVqO1cMpLEwuKbsuWjozik9sfB7XbdiJVFuiqPCvvU2YB7nJrRgXCdSrtVRMgqq3P10keoMWWAbKq8FX6fsRGIuhu3PLy6GtnZWyauu2t5NWvnT5AlrJyiSIUNsvIh8G8G+51x+E1YuTEEJKplYxKeW4U9wPIJ3rELl6YJmBBQUFekWA/qPl9wot1+rhtBbpXFLuumi25Usn0mzCPMhN22aVquoD3GQlEyA/PpZ1eQrWXXxKUU27h3f0VqQGn1PcVfJ+dFKKtbNSVm3TPB2pJEVaBQiS9fnnAFYC+C2s3p+XAigpwYAQQpw4syuf6Toncl/q7geWyXXYd+S3BRms/ekMBjNZ3HX5As/zsl2aJ3RtNmYVlmv1WL1sPhIxK1N05FAnBvsuRnY4BaWA7HCqqHF7EMI8yKuRpdjd04vOW7fi+K7NOL5rMxbcsrXo2umyCAXFyR5u1+ScVBJ7+9N46qV9uGRhB6a2jWWrtjTFsP2VA8bPrNqZi37XLGwGqW69AHBkaCRUhiszNqtLEIvaZwFcrZQ6CAAiMg3AF2AJOEIIGVc4g+fbkwkk4pJvXWVyHcpIKpTVa/PuzVj37BfRP/w61PQU4tll6O3vxOoHnwdQaBGphNVjUktTPk7NXYw2LAKEegCbXNxLz+jFkm+fi4Hh15HNpNB25AJ8+o8+FMgVvPqh5wvaifWnM0XXTmexNcWh9fanseCWrQVtynr709jw49cKatr1pzMFNfJsK9b2Vw7k+7am2hJoaYphIJ1Bqi0BpYDrSnRNugniog1j7bTXcstjLxZYUPvTGV/rnK537ZbffA9HJz2GWKIf7c3HINF+PQwdH0kIRPkUYxSRHqVUp99Y1Fi0aJHavn17vZdBCGkA7IdOb3+6yOqSiAkmtzah/2gGM2a9iNFpG5FR+drfaI23ov/Vi5DRiB8B8gVobTbv3lxUsV5lE3nLViqZwM6blxasTSd0gsR36fZ1EqanqH0+HzrzONy24hTP4+rW4XyoLz2jF4/uubvgOtqPIr/A88XrnzQKro5UUlt+I8i+5eC+Z5KJOC5Z2KF1nZYbl+e8V3X4XQMdputimkt3X7VNfR6tsx8p+r9hqs1HihGRHUqpRe7xIBa1mIhMdVnUguxHCCFVpVKlI5wPHfefrpmsQltzE3puWgpgOTbvLi6dcePuBA6iOKYr1ZYoGrv7ubsLRBpglchombkFI4c6i4qu6hrUi/hbabp7evGJjc8bOyOYEiNMPUXjIvj7laeVJDLccVlLH1pa8EAHkG9nNZB53bPdkJfFyM+atHrZfKx+8HlksuVWuytE5061W4i5x20ra6n3rrM8S6WSccK613VxkzLte0WfqV83AhKMIILr7wH8UEQegnU/rgSLzxJC6kylakAFKXbqfGDpWhd9Wunb/+g0kqk6uyn+DTA/nE3n7FUSw8ZUU80WjG50wf+lig2vCvWA9wPey4Xp5w5e0dlR5OarFqZrvzfX0cLpvu3tT2P1Q8Vuby8qmYwT1r2uE3CldCMgwQjSmeCbIrIdVqEgAXCxUupnVV8ZIYR4UG42pJ8LyYmfABjQtB4yjc+aNAt9R/qKxu0SGVM1VjibtZteDHTOQcSnV021Sb+3vsgdOmPWi1j6kK5vpV40eok40zVwYnrAr142vyhGDbBc1EGsSf0GkRbWDWyjS1AALAukTqzNSSVxy2MvFq0/M6pwy2MvhhJalSoQG7ZUjk7YldKNgAQjSNYnlFI/U0r9g1Lq/1KkEUKiQDnZkM4+o34EcSeFyWxcdfoqKE3fS7tExsGjGW0GaHdPr7YXJWCJJGcWYpBroAw9RQEg1twPEetn6+xHkJzVjcH2B9B3pA8KCn1H+vDQK3chkyyMA7ZFY3dPLz619T70T78Zk07qQv/0m/Gprfflz2nV6auQkBbP9eke8Lb4y4yqvKsUsKrz33lZMLes7jOx3cDu826a0qOdw9nL9ENnHqfNePzgu481ZkKaLHq1sPQ5sbOOr9uwE62JmLZ/qw5dlqc6cF7RZxqkGwHxJ5BQI4SQqFFO2Qc/i5OtAfweWDZe5QncJTgyAwuQHPiAZ4kMXYN6v7ILzsb27UmzVc4+rzvO7UJrvLX43F093yWWQTz1I0BTP65l5pai/ff2p3H79+9H7JiHCoRP7JiHcPv37wdguY8/u+QWtCeOAVSxi1j3gHeLa6Wsa/ylyxdg581LA1uWdJ+VlxvYjR1g/+v1y7F62Xw89dI+pDOjiOcunH3P3LbiFG2Der91ejV/D1LOJSi2+9UuKXPwaAZHhkd8S8oAhQ3u7XP73NKr8dklt2D2pNkQCGZPms1Eggrhm/XZqDDrk5DxTTnZkCd0bTY2T+8oIynB7eoDoF2jLiNQRyqZwKSWJuzNPUyDousj6jy+fY02795ckBhhckcqVSzg7PHDL60vGOtIJdE//WbENG6w7HAKL37s6aJx9zqcvU1twmYmeuH+rA7NvhY6B6b7/JzXrpz7b8EtW43WUdM8XscDwseqdd66VWvBm9qWyCXOkFpjyvqkUCOENCylBrNX8qFf6nFWL5tf0DA9LF4xVQLgrssX5GPw7HgpPxG69KGl+vg5JRApflZkh1M48quu/GtbOHzm+T8pqD82NhHwwp+9oHnDH5O41pVACYvpvDEyFfE9N2pbf5VzD3X39Ppmn9rz+MVS6kR5EMF4fNdmz2NXo0sI8aac8hyEEBJJSg2mrkWfUcA/jq7Uhtp+pTXm5Fr3hL02q05fpa3xlulfiERqh7b1VFwEWaUKhMwXfn4MBjKvF83f3nxMSecLVKbwr0nY6867Nd6Kte/9JJbP01uXyomRdGZsmgSYnR3qV+C2lH6sfi7TUjOoSXWgUCOERI5K1Efzohp9RnVr9hIXQTIzTXjFVCXSi/KCM+x1tN2Ndz93N/oO9yHrsNSNpt+mteAJVJFFa82Z1+Mz224uqKuVkBasOfP6ks4X0NdAC5rpCfiVcxk7by/3q5NyhaMtpE2WuXLvES/BGKTNlE7sVfv/JdFDoUYIiRSVqo/mh5fFKewDybRmU3X61cvm47oNO43zdaSS6D86jCPD+oe0qbRGLNGP23Murxu7X8D9z76adxcGvY52nTi3q9HUekonTJyCL6jwCYTbnapzrxrwK+eiq4/nRaWssl7zeN0j9nYtTTGt+9xLMAZtM+Xcrlb/L0kxzPokhEQKrwdqqYTJlnNmFzozKb32Ma35qZf2GTP/TA9SOz5o2MMtaiqtMXvy7Hygu1OkOdcU9DoGsQx5CZPl85Zj66VbsevqXdh66dZQIkj3edllOZxkRlXg8ym3ub0bXeZjKe2hvObx+gziIrhkYQfWXnhy6IboQa1+zu1M9/gnNj5fVvZprahkxmytoUWNEBIpKv1ADWsJKKWQrteaTZY7L0vKnVte9gw0H9q3rCBGDSgsaXHnlpeNWaJBe12aGoDbBV6rFXBu+rxMLsCg90UlYtzc+MUBBrXMhrlHbEaVwsM7erHobdOw7uJTQlmAgzR3d4s903UeVSrylrVGtwZSqBFCIkWlH6hhhVcpQrGUNdvHdrY0EqhAmaAjhzoxCKD1mC2QxABmu1yLXmsVWA8uvweUbn2AJdKSiTjOPmkm7tzyMq7bsLOi8Uqmz8ur0r8TkzgqxVXZ3dNb8HlMbUvg5gtODhS3VQlx4Jd0YN/HfnXPvOa11332STPx1Ev7Cl47P99UW8JYkDdMR5B6UG4Xk3pDoUYIiRSVzsgMK7xKEV3lrHkwM+biPJrJ4mjG7PKc2pZAW7NVV+2tsfdg9cKPaB80Xv0wFRD4AbWiswN3bnm56AGdzoyWFP8WBC/LTTIR97zGQcRRUMuTroTGwaOZgp6cXserlDiwrW2m8iTluG69zt19XomYIBGXIvdzueuoBZW20tcaCjVCSKSodEZmWOFViugKsmad5SVMVl8yES+y5pjwc22FeUCZtjXFv5Ur1EyfV4fjmpmusZ84ClOyxOR+tuPibBFrOl6lxUE1XLcmdOeVySqkkgm8OThSlmWzHoS5dlFatw2FGiEkclSq2TQQXniVKhTDWij8YoSchI0Hs7f7xMbnAz1UvfCyzrmphIXC6/Pyuy/KFUfOh7RXKXh7PtN16e1Po6PCwqpWtf8A8/UaSGdw1+ULfNcRtZiwoNcuauu2oVAjhIxrShFelRSK9rHDxF05iYuU1C3BXn/Qh7upjZPuIWcnFLiphHWnHItqOVanIMVlbVJtVi9V0+cXF6m4sKpG7T8TXtcxyDqiFhMW9NpFbd02FGqEkHFPpYVXGLp7eo2WF13clW4b07x+Dx7TAwqwWiDZY0vP6MV3996Tr8zfd6QPa3+4NjfH8qI5zj5pprE+XCWoR8eJMG7ow4Mj6O7pNX42o0pVRVjV6j72u47VtmxWgyDXLorrBijUCCGkathWGhPOuCuTmOswxNEEddG4H1C6fR/cfS8kMViw3+DoIO5+7m4sn7dc+5Bb9LZpkYvlKUcchXkYZ7JWnJrJvWl/ZkGFVZCm9LWkXJFZzXi6asaQ1TIOMAwUaoQQUiW8rDTuuCud681kDSrHRaNdU5O+00Hfkd8a56mnldKLUtcVJhYPsIRdkHgtPzbv3lzQZ9Rpzay3WCv1861WPF21Y8hqGQcYBgo1QgipEnv702ia0qPtkemuYu+2YrQnExABrtuwE3duebmgzpVXMVunS1NnbdBZjlQmBWkuFmsq0x6o5lqlqFXGne44QYrAOgkar+XH3c/dXdAMHii0ZjYi1Yqnq3YMWS3jAMMgyieQtVFZtGiR2r59e72XQQiZwLzrS3cg3f5AQQcBlU0gOfAB/OTaG4z7hQlsd+IO8k8m4kWCUNcEvGlKT1GnA5VNYLDvYrw19p6SkhnCYrIoltKWKexxACCVTOD802YXFX3dvKuvqI5cJdd16n2nQmmkt0Cw6+pdZc8/njDVkhMAv17fmKLWiYjsUEotco+z1ychpGFotH59LcdsKRA/ACCxDFqO2eK5X5jA9vy8MNc2c7J62fyi3pAjhzox2HcxssMpKAVkh1MY7LsYI4c6axZIXY0er0GPAwD96Qwe3tGL1cvm49frl+OZrnNw24pT0HPTUnzp8gVl9/Q0MWvSrFDjExlTrFi9Y8iqDV2fhJCGIKo1jrw4lNkXatwmjDgSeMdXuecy1VgbOdSJkUOdRft7PQQrGQRfq4w7r/lMbrRy4/G8XLqrTl9VEKMGFPZtJWNENYas2lCoEUIagnLiU6oZ++Q196xJs9B3pK9oHz9rSdDA9o5UMu+W1Lk07bncmGqsufF6CFY6CL5WGXd+1zZMQkEQ/P7AsK9VlLI+o0pUY8iqDV2fhJCGoFSLi/2g7M0F4dsPykq4Tf3mXnX6KrTGWwv2aY23YvG0Kz1duDr3pBu3iNLt49eBYd3FpxS49D585nGBXXxeQfBhsN3Zvf1piM85VoLVy+YjEXcfqXr4uXS7e3rxuY1J/GL7Kryl70v469/7BkWaBys6O/BM1zl59/R4F2kALWqEkAahVItLNTPF/ObWWUsWT7sSDzw1E+nMWBsitwtXZzlwZn3qLAm17sDwW0PpDtO4Dre1SWEs1i5s26xQ1DCHzusPjEZ055PaQ6FGCGkISo1PqWbsU5C5l89bXmAhWbz+ybxIs9EJx1JEVC1rm5ncuhhJBS7poRO6tkirVqapqdm6TVwqa23z+gMjqi2LSLSg65MQ0hDoXHVBsu+qmSlWytxBhWPUM1x1bl2VTeDo75YGdi3Xo2WP39wffPexFT2el0u6Vucf9XuJeEOhRghpGEqJTwkbuxWGUuYOIu6qGVdXKZbPW46171kLGZlaVNIjaFmNepRb8Jo7HhMsetu0ih7P6w+MWpx/I9xLxBsKNULIuKZUS1y15j77pJm+QfO1qilWLsvnLcebv/gkDr+0Hkd+1VVQ3iOIVaiaIjrMMW1Gcz08K43pD4xanH+j3EvEDGPUCCHjnmrGboWZu7unFw/v6C2IZRcAlywsnKMeLsFSKaesRj3KLdhzX7thp/b9Wl7jWpx/I91LRA+FGiGE1AhT8PxTLxUWwK1VTbFKUG4R0no0d1/R2YE7t7xc8jWuZF2+ap9/I91LRA9dn4QQUkG8AreDWjfq4RIslWq6lqtJqde40WK+GuleInpoUSOEjDuq2YnA77hedbGCWjeCusTqdZ5uamkVq9Q5l+p2bLSSGhO1mv94gkKNEDKuqGcRUb+H+NknzcS3nn21aL+zT5pZNOYnfiZisdRKn3MpArMRY77q4V4mlYOuT0LIuKKeWW5+D3F3LJqNadwLr/PcvHszlj60FKfedyqWPrQUm3dvDj1/FKnEZ1tuTTFTbJcCWKMsAKzpFh5a1Agh44p6Wjz8XJumht+lrM20z+vZH2LtDx+tWLP0KFHuZ1sJi5wuecKmElbNqLizq7GOiWgFrgS0qBFCxhX1KKJq4xW43d3TW1Q/rZy1mfZJvnVrRZqlR5FyP9tKWOScyRM6yrHeRiVRoVrrYE230qBQI4SMK+qZ5eaVAXnnlpe1vcAlt+awmM5TNfVrtw/TLD2qlPvZVsraahewNQnvUq23UREy1VpHI8b3RYGquT5F5FgA3wQwC0AWwL1KqbtFZBqADQCOB/AbACuVUgdFRADcDeBPARwF8GdKqedyc10N4Mbc1Lcppe6r1roJIY1NvbPcTIHbpoeRQumB8EDxeX75V/pm6bMmzQp9jKhR7mdb6ZpilZ6vkq7xcqiWoGJNt9KoZozaCIBPKKWeE5G3ANghIo8D+DMATyil1otIF4AuAJ8EcB6AE3P/3g3gKwDenRN2NwNYBOs7bYeIbFJKHazi2gkhDUwUs9xMDymTCy0IuvNMtK/C2h+uLXB/tsZbser0VSUfJ0qU89mWW5y3mvPZrnGd1bXWQqZagqrS13+iUDXXp1Kqz7aIKaXeBPBzAB0ALgJgW8TuA7Ai9/tFAL6pLJ4FkBKR2QCWAXhcKXUgJ84eB/An1Vo3IYRUg1q5ZO1m6bMnzYZAMHvSbKx9z9qGTySoBJUuzlvJ+SrtGi+Hat2rjVocud6IUrpbo8IHETkewA8AvBPAq0qplOO9g0qpqSLyXQDrlVLbcuNPwLK0/TGAVqXUbbnxzwBIK6W+oDnOxwF8HACOO+64ha+88ko1T4sQQkIRlYw+Ej1O6NqsFWqAJWhqfa/wXq09IrJDKbXIPV718hwiMhnAwwCuVUodskLR9JtqxpTHePGgUvcCuBcAFi1aVH0FSgghIYiiS5ZEA5O7EahPGQveq9ERq1XN+hSRBCyRdr9S6pHc8O9yLk3kfr6eG98D4FjH7nMB7PUYJ4QQQsYFOnejE5axqC1RKZUCVFGo5bI4vwbg50qpLzre2gTg6tzvVwN41DF+lVicCWBAKdUHYAuApSIyVUSmAliaGyOEEELGBX712QCWsaglUSmVAlTX9bkYwJUAXhCRnbmxTwFYD2CjiHwUwKsALsu99++wSnP8ElZ5jo8AgFLqgIh8FsBPctvdqpQ6UMV1E0IIITXHdjcuXv8ky1hoqKUrMko136om1HJJAaaAtHM12ysAf2OY6+sAvl651RFCSGMTlfgZUnlYxqKYWrefilLNN3YmIISQBiNK8TOk8rCMRTG1dkXWs8OJGzZlJ4SQBsProTWRH+bjCWZdFmJyOZoyZcul3h1OnIYgTXoAAB2+SURBVFCoEUKIhii7FhcdehwbmjdijuzHXjUDd4ysxKbsEgabk3GLyRUpsP6vVuP/ZlTEMl2fhBDiItKuxV0bsb75a5gb24+YAHNj+7E+8VVcGNs24YPNw9Dd04vF65/ECV2bsXj9k9H4bImR1cvmG4uqjveyJRRqhBDiIkqp+UU8cSuSGCoYapNhfDKxcUIHm4ch0kKcaFnR2WHs3DDeLckUaoQQ4iJKqflFDOzRDs+RNyLhpmkEIi3EiRFTjbnxbkmmUCOEEBemL/5IPBDa52qHxTBOiom0ECdGopSJWUso1AghxEWkHwjn3gQkXIIxkbTGSSAiLcSJkYlatoRZn4QQ4iJKqflFnLrS+vnErZYbtH2uJdLsceILC8o2LlHJxKwlYjUEGH8sWrRIbd++vd7LIIQQEkGiXH6FTExEZIdSapF7nBY1QgghE46JaJkhjQlj1AghhBBCIgotapVm10bGjhBCfKHrjRASBAq1SrJrI/DYNUAml+I98Jr1GqBYqyQUw6TBsQuu2sHsdsFVABRrhJAC6PqsJE/cOibSbDJpa5xUBlsMD7wGQI2J4V0b670yQgLDgquEkKBQqFUSQ8Vw43gYdm0E7nonsDZl/YyKMKn1uiiGyTiABVcJIUGh67OStM/NWXo04+UQNZdq3vX4GgAB7A5stVhXNcUwITViTiqJXo0oY8FVQogbWtQqSZiK4WEsUWGtSNW0chW4HgG42+RW27plEr1sn0MaiEh3PiCERAoKtUpy6krggnuA9mMBiPXzgnuKrUth46zCWJFKjeEKKu50otFrXZUWjToxDLHOM0ouYUI8mKitcAgh4aHrs9KcutLf7edlIdPtG8alGnZuIJxrNYiL0V5XNVy2Be1z6uB6JaRCVLvgKst/EDI+oEWtHoSNszpxKSxB4sDkUi0lhiuMa9XPxehcl2ne732ycCys1e3UlcB1P81ZLmvseiWkAbDLf/T2p6EwVv6ju6e33ksjhISEFrVas2sjIDFAjRa/pxNBuzYCz38bRkHS8y3gN9us+SQONLcBw0eCzW1jFHevWeLJWavs3JsKrWQA8lat9mMLa5qZ5k0fAD5/ApA+CCSnAsOHgdHhsWMGtYoxsYAQLV7lP2hVI6SxoFCrJbYrUCfSTBYyr5iwgdcKXaJq1BJpsTiQdRzDNLeNybVqTWq91/3X1ssC16NPwVmvedMHCn868XPV+s3PxAIywWH5D0LGDxRqtcRLdDlddk6BUop1yCnSktOA8z7vLXq0VjL3nBnLZWnH4HnFu33vk3oBFoYg561bt58oJWQCwPIfhIwfGKNWS/zEx8BrwCMfA9a2j8VqlWsdSh+whJNX3Jc7W9VrLhtdXNmujZblrVyRBgQ776BZtoRMMFj+g5Dxgyil/LdqQBYtWqS2b99e72UUctc7PVyMOhwZjeWSSOInp9yCa392on8W2Np28zxrB4DvXg9s/3rh2hJJoClZGZGWSFJwEVImzPokpLEQkR1KqUXucbo+a8mJS4sFjicVFNGZNObsuAO9Q/cAABYeehxndV8F9eiblg3N6SKVGKCyxXNIzLKa6c4hk/avrwaxkgeG3rRcqSZOu4IijZAyqXb5D0JIbaBQKxPfv1pN7ZbqwGzsBwBcGNuGLyTuRbOMjL2ZPmC5LV99Vi/SAGv8iVtR0jm0H2uV1ACsa/Kdv9InVQDAL7Y6rptPwgIhhBAyjqHrswzsWkXpzCgujG3DDU0bMUfewGDbLLSdd6slekJZ0KqLUkCvmoE2GcQ0ORx+gvZjc3F2JZzPxf9cKLTWprznSSSLkwRsdyhFHCGEkHEGXZ9VwK5VdGFsG9Ynvoo2sWqBtaX7LOuUl3uvDogAc2U/StPmYgmivHUwJG4h5VkSBN4FeKPUoJ4QQgipIsz6LAO7JtENTRvzIi1PxESaE/FI7DShoHD0ezdZcXaxRLid248tHiulhIadFRumQT0hhBDSwFColYFdk2iO7K/zSqqPwLIUZrd/DYAhhs3EiUsLX9uuS+PB4ub3TLAbASGEkHEIhVoZ2LWK9qoZ9V5KzYgBhQV1g/CLrWO/290ZvNye00+EZz03HexGQAghZBxCoVYGKzo7sO7iU/DV5g/jqGqu93Kiy8BrY0Vxvboz2Ox/GaESFuLNVuusoE3dCSGEkAaBWZ+VoqAMB6kZyWnFddlYMJcQQkiDYcr6pEWtUpy6cqxOGKk+iaRV8qN5UnHiBpMLCCGEjBMo1CoJXW7VIzlN39PTlETA5AJCCCHjANZRqxR2kDypAmJ1TmieBFx8b6FL01SPjckFhBBCxgG0qFWKIEHyJDj5Eh2Otlt2cVun5fLcmyw3qJNEsrQ6bYQQQkjEoEWtUtDVVkFiQGu7ZUXTNX//zl9Z7bl+sdW67smpQFMSSB9kSylCCCHjCgq1cskXbx2f2bP1IZsTaQbUKLD9a2Ov0wdyyQX3UqARQggZV9D1WQ5BireS2sBMT0IIIeMQCrVyYFxatKD7mRBCyDiDQq0cKAyiBTM9CSGEjDMo1MqBwiA6MNOTEELIOIRCrRzOvQmIJeq9ioZCKetfWXMAwAl/pC+ASwghhIwjmPVZNtl6L6BiKAWIhH/Pb06bA2oybhm5CgCwNvFNTMVhAMBh1YLJMhRofqWAb46+D+2n3oMVnR0F73X39OLOLS9jb38ac1JJrF42v2gbQgghpJGoqlATka8DOB/A60qpd+bGpgHYAOB4AL8BsFIpdVBEBMDdAP4UwFEAf6aUei63z9UAbsxNe5tS6r5qrjsw3/skkB2t9ypCYxJdCsCoiqFJisVnFoK4RwkS85yCeUP3F41vGlpS8Hp3yxUIogOHEceO7DuwY8vLAJAXZqm2BA4PjiCTtdbY25/GdRt2YvsrB3DbilMCzEwIIYREj2q7Pv8FwJ+4xroAPKGUOhHAE7nXAHAegBNz/z4O4CtAXtjdDODdAM4AcLOITK3yuoPhVesrwpgsV3vVDFyf+SscVc0F40dVM8SnTpxpzoNqUsHrC2PbsK35GuxuuQLbmq/BhbFt+WMHoUVGcUPTRvT2p7H6wefR25+GAnDwaCYv0mwUgPuffRXdPb2B5iaEEEKiRlWFmlLqBwDcauYiALZF7D4AKxzj31QWzwJIichsAMsAPK6UOqCUOgjgcRSLP1ImSgF3jKzEpuwSdGX+AnuyM5BVwIiKIYlhZEu8VZwC7sLYNqxPfBVzY/sRE2BubD/WJ76KC2PbcMfIyiKBaKJD9mNb8zU4D0/7nxcsqxshhBDSiNQjmeCtSqk+AMj9PCY33gHAWTl2T27MNF6EiHxcRLaLyPZ9+/ZVfOFFJKdV/xg14iAmY1PWckduyi7BHSMrMYhmNEkWIkCTZEtKAkjhSP73G5o2ok2GC95vk2Hc0LSxQCCqnEA0HU9cIs+Pvf2sdUcIIaQxiVLWp855pjzGiweVulcptUgptWjmzJkVXZyW8z7fkFmfbgF0VDVjbeaqgjGdqLKtY0oBoypYZsFeNT3/+xzZr91mTs5C9qXElwEAqzJ/jbcPfQurMn+NIRXX7gOMiTw/YiI4oWszFq9/km5QQgghDUU9sj5/JyKzlVJ9Odfm67nxPQCOdWw3F8De3Pgfu8b/qwbr9McuB/HIx+q7jhAoBTydPRnz5HeYI29gr5qed3k6MYkqwBJsadWMmFIFYs6dUDCsmpDEIHa3XIG9agYOqsmYLoe1c86NWcebK5alDJncsXxSDObIG57vA8CoGkswWPPICwDAbFBCCCENQT0sapsAXJ37/WoAjzrGrxKLMwEM5FyjWwAsFZGpuSSCpbmxaHDqSkDMVp+ocQStuCrzaSwZvgfzhu7HkuF7ikQa4B/cPwlDeHD0rFwsm2BPdga+Ofq+/OsDajIUFKbHDufj0d4ig8hoLHEx15BtKbuhaSOaZcRzHU6LXRDSmVHGrBFCCGkYql2e499gWcNmiMgeWNmb6wFsFJGPAngVwGW5zf8dVmmOX8Iqz/ERAFBKHRCRzwL4SW67W5VS0Uq3XPhnwPav1XsVRegsXJ/K/HnRdhfGtuGGpo2YI/uxV83AHSMrccfISnwp8eUiEWUjAlwV/08cUJNxbeZ/58Xezbn3tzVfg2mxQutZs4wEdplaljLvoLijqhl3jIQvcsuYNUIIIY2CqHLLxEeURYsWqe3bt9fkWN09vbjo0T8IVAesligF9KoZni5OOxPT6cI8qprRlfkLLIz9D66M/6dRrNkMqThWZ/6yYO7dLVdo9wtaOHdP1rLo2S5R9xyjiCGGbF5Y6qyCJjpSSTzTdU7g7QkhhJBqIyI7lFKL3OPsTFAm3T29WP3Q8ziraTKmGeKvdJRa6T8MWQBLhu/x3MYrE3PJ8D3YkX0H7kp8GXGPtdq1zTYNj4mlvWoG5mri3EYRQ5NPN4esAp7ILsCO7DuKROSwaoKCQotYhYadMW06sSYotMslE3GsXjbf8/iEEEJIVIhS1mdDcueWl5EZVaFKV2QV8JLqQNan76XfnH59MxVivuUrzJmYVpD+puwSXJf5a98aZ+6gfl1dtKOqGfePnuN7XjEBLov/AAAcNd2sOLg3VWtepNl4ZX8qWBY0yf1cd/EpTCQghBDSMNCiViZ2vNNUgzVNKUssZBFDHFn0qhl4IrsAl8V/4OtSNGFb45zlMnTWuSbJWgJmpLC3prPnZjbnQiw6L0eQ/qbsEiADfDHxT9r2Uu7tnftYsW9jrlfAim3zw2nVc1rqdrdcod3elP05tS1BNychhJCGhUKtTOakkujtTxtdfb1qRpH7cVvzNUXuRh0m16h73MuF2iH78YXEvQXZk9PlMP4+8RUoxLXCSxekbwsv91yAFaPmHdQ/ZkK7oWljYJevTnyZrrMp+3OchmASQgiZIND1WSarl81HIi5aV9+QiqNNBov6WnrVKPMjrPAYRUxb4iIhSjs+omLoyvyFNt5rU3YJ/i7zcav0Rs7t+kZ2clEiAWBuFxXm3HXiy+RSNQnFgXQm8PEIIYSQqEGLWpnY8U6f/k4MXQ5X30E1CW+RwXyCgTPo3WQVcpJVxfXFwqIUtG5NL2JQBaLLWbqjH5ZAS+Ewen2yLU1JCiNK72p1u29N4svtUu3DdOxdeAN2/OxEQFN2Y04qGfTUCSGEkMjB8hwVpPPWrTh41LLgbGu+RltaYk/WEjjubMYhFccRJJHCEexV07FbvRVLYi8WiDVTLJppPKusrMcw2aV7smOuWl3pDidHVTMeHD0L58Z2FtRg25Rdgl+3XGFc6zDiBQkBhfOYS4nYJBMxDGaymJNKYvWy+VjR2YHunl6seeQFpDOjju3iTB4ghBDSELA8Rw3oPzrmZvPKpjQF2tvCxBZIQUTaiIrhh9nfx3tjLxa8r3wschklUIgXuD/dViydVcxJmwzjqvh/5o/rtBqaynCIAKIEb2QnY6ocKTj3m4u2LqTDIczc2GN3bnkZe/vTaE8mIAJct2En7tzysnE/QgghJMpQqFUQO7EA8A9635RdUpDNCIy5GTtkf+CEgRgUrsp82uGifANZiDZJwDaeOrM+TWIRCBZL516Xna3p5XJtlhGkVSveNXwvsgENukGK1K7o7NBa19jjkxBCSKNCoVZBVi+bnxcIOvemV9C7n5vRhE74mUpYKAjmDd1fMOYWi4Vz+8fS6bCEn/e+c2Nv4IsrF+DaDTt95wtbpPbOLS8XuECBsR6fFGqEEEIaCWZ9VpAVnR1Yd/Ep6Egl8Vh2Cda4irWasikBfzcjgCLrk0n4mRqqh21grsuwDIJtnfPad092Om557EXPeUotUmvq5cken4QQQhoNWtQqjO1+u7H7BXzr2SV41MNi5cTPzRgm4D6sNc+EO5auH5OgFDBNDhtdsfZx7H3tQru6jM6Dw9UpneF0QbvHCSGEkEaCQq0KdPf04lvPvhpqH5Or0G5AnsQwzo3tDNSA3KsrwLbma4oyNP3mcrpHL4xts4reorAGm1LAQUzG2sxV+Tk3ZZdg09ASXNr8Q/ydbMAxar9vRmfBnCgtvszpgrZhj09CCCGNCMtzVIHF65/UWnS80MWoDak4BFKUmenlQg0zfylzmcqOHFCTcfrQvaHW5CQuglGPezFIMoGT7p7efAboHI9sUUIIISQKsDxHDSklFkpnBWtzFMy1sbMqvZIAdJgK0IaZS2B20aZwJNR6nExtS6DnpqUAgBO6NkMn18JeU9sFTQghhDQyFGpVwBQj5YfbzRi2AbnnmjzqugVFIXyvTT8ujG1DV3YjsPYNoH0urp58Cf7l8BnF62R8GSGEkAkIsz6rwOpl833bP10Y24ZtzdcU9QF10ofKZG9a+5jnSsQEiXiw9gVhe216Ybtj52A/AAUMvIYb1T/h0uYfFmzH+DJCCCETFQq1KrCiswPtyYT2valtCayIP6NtWO4Wa5/PrEQa5YuiGIAvZC/XCqyvNn8Yd152Gu689DR0BLBabcouQVeIsiNe6NyxTaODuHXSw+hIJYOX59i1EbjrncDalPVz18bQayGEEEKiCJMJqoQp1koAPO3RB9Tus2lzUWwb1k56GO3DrwfOmIxJcc01AAXdC/aq6fhq84ex9sZbCrYpJRGiVHa3XGGwPAqwtj9YQsCujcBj1wAZx5oTSeCCe4BTw1v5CCGEkHpgSiagRa1KmGKq5qSSmBPTx4Xp4sW2T3k/pn76f7Bj4efRFBN8KfFlPNu6Ch9sfdZ4bFNbpk3ZJVgyfA/mDd2PJcP34D5NLNjqZfORTMSNc7tJJuKY2qa3HgKWRezDZx6Xt5ClkglMbUtAALwuM/U7SQw/2fT/sOaRF9Dbny4o09Hd01u47RO3Foo0wHr9xK2Bz4EQQgiJKkwmqBJetbwGt85CW7qvaJ8+FMae5WOzdm3Eu3auATAKCDAL+3C7/CMyzVk8NPye/PZjFrNgddJ0YtK2WO3cfC/+YvhbnnPZTdIB5M/VabUbbJuFtvNuBU41lNXYdaTYGgYAahTvfO4zeP/oR7EJY8fUtoEa2KOf2zROCCGENBAUalXCFhNa1138Vow8+n/QNDqY334k3oq9p92Ajp8li7f/3LVAtrB3ZUyN4s7YP2I4lsWm7JKiOmlzxYp7Q8aypAlQ4Ir1CtBfEX8GK+T/AbG0dq5UMoGdNy8t2m/n5ntxQ2ZsDW3pPkuIAXo3pD32nb8EVGET9ySGtKVDisp0tM8FBl4rnrt9rvbcCCGEkEaCMWr1YtdGyz03sMcSFScuBX6xdez1uTeNCZm17cZphhHHYZUsatNksyc7A+9X/4hLFnbgqZf2BSsAe9c7teLHnssY3G/YD+3HAtf9VH+sXRuBRz6mfSuripvIFxW+ZYwaIYSQcQAL3kaNU1eOCQm32Bh4zdsS5aAZo0VFcZ3Mib2BdReFa2puchvOie33nksn0rzGAc9YMqMr2Il9fZyi1ylyCSGEkAaGQi0KeAXEn7oSKHJcBifWPjd8hX6DOzEGwYr4MwAMIkjigBrVj5vwiCXbu9DgCnbjFL2EEELIOIJCLQr4BcQv+nNg+9fCz5tIWtalsJx7E/DIx1EsDpVDPGrQiTSvccAcY5achndd+Jd45sIgCyaEEELGJyzPEQVMge/2+PlfBBZ9FJZlLeicx5Yep3XqShgteF7ZlO3HhhsHLFGYcGWfJpLAeZ/3XCIhhBAyEaBQiwImseK0hp3/RWBtv79gSySBi//ZCt4vxx1oFF0e2ZRBzsPNqSstQdl+LAApT2ASQggh4wy6PqNAmID4X2yF0drVfmzlAunPvUmfTeknuoDwgf2MMSOEEEK0sDxHo7E2Bb1Qs9ouVRR3CRFmUxJCCCFVgeU5xgu1LPBKSxchhBBSVxij1miUEgdGCCGEkIaEQq3RYPA9IYQQMmGg67MRoUuSEEIImRDQokYIIYQQElEo1AghhBBCIgqFGiGEEEJIRKFQI4QQQgiJKBRqhBBCCCERhUKNEEIIISSiUKgRQgghhEQUCjVCCCGEkIhCoUYIIYQQElEo1AghhBBCIgqFGiGEEEJIRKFQI4QQQgiJKBRqhBBCCCERhUKNEEIIISSiUKgRQgghhEQUUUrVew1VQUT2AXil3usIwQwA++u9iAkCr3Vt4HWuHbzWtYHXuXZMxGv9NqXUTPfguBVqjYaIbFdKLar3OiYCvNa1gde5dvBa1wZe59rBaz0GXZ+EEEIIIRGFQo0QQgghJKJQqEWHe+u9gAkEr3Vt4HWuHbzWtYHXuXbwWudgjBohhBBCSEShRY0QQgghJKJQqBFCCCGERBQKtTogIseKyFMi8nMReVFEVuXGp4nI4yLyi9zPqfVeayMjIq0i8mMReT53nW/JjZ8gIj/KXecNItJc77WOF0QkLiI9IvLd3Gte6wojIr8RkRdEZKeIbM+N8bujCohISkQeEpGXct/Xf8hrXVlEZH7uXrb/HRKRa3mdx6BQqw8jAD6hlPp9AGcC+BsR+QMAXQCeUEqdCOCJ3GtSOkMAzlFKnQZgAYA/EZEzAXwewF2563wQwEfruMbxxioAP3e85rWuDmcrpRY46kzxu6M63A3gP5RSJwE4Dda9zWtdQZRSL+fu5QUAFgI4CuA74HXOQ6FWB5RSfUqp53K/vwnrP38HgIsA3Jfb7D4AK+qzwvGBsjice5nI/VMAzgHwUG6c17lCiMhcAMsBfDX3WsBrXSv43VFhRGQKgLMAfA0AlFLDSql+8FpXk3MB/Eop9Qp4nfNQqNUZETkeQCeAHwF4q1KqD7DEHIBj6rey8UHOFbcTwOsAHgfwKwD9SqmR3CZ7YIlkUj5fAnADgGzu9XTwWlcDBWCriOwQkY/nxvjdUXnmAdgH4Bs5d/5XRWQSeK2ryQcA/Fvud17nHBRqdUREJgN4GMC1SqlD9V7PeEQpNZozqc8FcAaA39dtVttVjT9E5HwAryuldjiHNZvyWpfPYqXU6QDOgxU2cVa9FzROaQJwOoCvKKU6ARzBBHa/VZtc/OqFAB6s91qiBoVanRCRBCyRdr9S6pHc8O9EZHbu/dmwrECkAuRcFv8FKyYwJSJNubfmAthbr3WNIxYDuFBEfgPgAVguzy+B17riKKX25n6+DiuW5wzwu6Ma7AGwRyn1o9zrh2AJN17r6nAegOeUUr/LveZ1zkGhVgdysTtfA/BzpdQXHW9tAnB17verATxa67WNJ0Rkpoikcr8nAbwPVjzgUwAuzW3G61wBlFJrlFJzlVLHw3JfPKmU+hB4rSuKiEwSkbfYvwNYCuCn4HdHxVFK/RbAayIyPzd0LoCfgde6WnwQY25PgNc5DzsT1AERWQLgaQAvYCye51Ow4tQ2AjgOwKsALlNKHajLIscBInIqrCDUOKw/SjYqpW4VkXmwrD7TAPQA+LBSaqh+Kx1fiMgfA/g7pdT5vNaVJXc9v5N72QTg20qp20VkOvjdUXFEZAGs5JhmALsBfAS57xLwWlcMEWkD8BqAeUqpgdwY7+kcFGqEEEIIIRGFrk9CCCGEkIhCoUYIIYQQElEo1AghhBBCIgqFGiGEEEJIRKFQI4QQQgiJKBRqhJAJi4j80DD+LyJyqe49w/bHi8hPK7cyQgixoFAjhExYlFLvqfcaCCHECwo1QsiERUQO536KiPyDiPxMRDbD0QBaRG4SkZ+IyE9F5N5cZxGIyEIReV5E/hvA3zi2j4vInbl9donIX9b6vAgh4wcKNUIIAf4XgPkATgHwMQBOS9s/KKXepZR6J4AkgPNz498AcI1S6g9dc30UwIBS6l0A3gXgYyJyQlVXTwgZt1CoEUIIcBaAf1NKjeaanj/peO9sEfmRiLwAq9n8ySLSDiCllPp+bpt/dWy/FMBVIrITVlu46QBOrP4pEELGI031XgAhhESEon56ItIK4MsAFimlXhORtQBaAYhue3s3AP9HKbWlWgslhEwcaFEjhBDgBwA+kIsvmw3g7Nx4a+7nfhGZDOBSAFBK9QMYEJElufc/5JhrC4D/LSIJABCRd4jIpKqfASFkXEKLGiGEAN+B5dZ8AcD/APg+YAkyEfnn3PhvAPzEsc9HAHxdRI7CEmc2XwVwPIDncokH+wCsqPL6CSHjFFHKZL0nhBBCCCH1hK5PQgghhJCIQqFGCCGEEBJRKNQIIYQQQiIKhRohhBBCSEShUCOEEEIIiSgUaoQQQgghEYVCjRBCCCEkovx/FDz1unVQPb8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot('idade', 'conta_corrente', 'km_3')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cluster 0: risco bom = 0.0 %\n",
      "cluster 1: risco bom = 100.0 %\n",
      "cluster 2: risco bom = 100.0 %\n"
     ]
    }
   ],
   "source": [
    "for c in sorted(df.km_3.unique()):\n",
    "    n = df[df.km_3 == c].risco_bom.mean()\n",
    "    print(f'cluster {c}: risco bom = {n * 100} %')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cluster_report(df, columns, cluster_column):\n",
    "    for c in sorted(df[cluster_column].unique()):\n",
    "        print(f'Cluster {c}')\n",
    "        for column in columns:\n",
    "            if len(df[column].unique()) > 2:\n",
    "                print(f'- {column}: mean = {df[df[cluster_column] == c][column].mean()}')\n",
    "            else:\n",
    "                n = df[df[cluster_column] == c][column].mean()\n",
    "                print(f'- {column}: count = {n * 100} %')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cluster 0\n",
      "- idade: mean = 32.95122625257916\n",
      "- conta_corrente: mean = 2095.40948161314\n",
      "- risco_bom: count = 0.0 %\n",
      "Cluster 1\n",
      "- idade: mean = 22.35180574907043\n",
      "- conta_corrente: mean = 1112.2331189415725\n",
      "- risco_bom: count = 100.0 %\n",
      "Cluster 2\n",
      "- idade: mean = 36.978460955791476\n",
      "- conta_corrente: mean = 4808.145796072836\n",
      "- risco_bom: count = 100.0 %\n"
     ]
    }
   ],
   "source": [
    "cluster_report(df, df.columns.values[0:3], 'km_3')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['km_4'] = KMeans(n_clusters=4, random_state=1).fit_predict(X)\n",
    "df['km_5'] = KMeans(n_clusters=5, random_state=1).fit_predict(X)\n",
    "df['km_6'] = KMeans(n_clusters=6, random_state=1).fit_predict(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cluster 0\n",
      "- idade: mean = 23.303504116727744\n",
      "- conta_corrente: mean = 892.8037889276358\n",
      "- risco_bom: count = 0.0 %\n",
      "Cluster 1\n",
      "- idade: mean = 22.304747860872794\n",
      "- conta_corrente: mean = 1111.0541513348442\n",
      "- risco_bom: count = 100.0 %\n",
      "Cluster 2\n",
      "- idade: mean = 32.453925657389604\n",
      "- conta_corrente: mean = 5153.194669563995\n",
      "- risco_bom: count = 100.0 %\n",
      "Cluster 3\n",
      "- idade: mean = 54.14411020347421\n",
      "- conta_corrente: mean = 2746.411343282708\n",
      "- risco_bom: count = 0.0 %\n",
      "Cluster 4\n",
      "- idade: mean = 56.934335179649985\n",
      "- conta_corrente: mean = 3272.9654266867205\n",
      "- risco_bom: count = 100.0 %\n",
      "Cluster 5\n",
      "- idade: mean = 36.82627224263472\n",
      "- conta_corrente: mean = 4577.569291749365\n",
      "- risco_bom: count = 0.0 %\n"
     ]
    }
   ],
   "source": [
    "cluster_report(df, df.columns.values[0:3], 'km_6')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "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>idade</th>\n",
       "      <th>conta_corrente</th>\n",
       "      <th>sexo</th>\n",
       "      <th>risco</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>21.257389</td>\n",
       "      <td>783.127911</td>\n",
       "      <td>masculino</td>\n",
       "      <td>ruim</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>21.766573</td>\n",
       "      <td>979.747521</td>\n",
       "      <td>feminino</td>\n",
       "      <td>bom</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>25.321033</td>\n",
       "      <td>1065.328054</td>\n",
       "      <td>feminino</td>\n",
       "      <td>bom</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>23.919268</td>\n",
       "      <td>1195.758078</td>\n",
       "      <td>feminino</td>\n",
       "      <td>bom</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>21.805298</td>\n",
       "      <td>1083.764450</td>\n",
       "      <td>feminino</td>\n",
       "      <td>bom</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       idade  conta_corrente       sexo risco\n",
       "0  21.257389      783.127911  masculino  ruim\n",
       "1  21.766573      979.747521   feminino   bom\n",
       "2  25.321033     1065.328054   feminino   bom\n",
       "3  23.919268     1195.758078   feminino   bom\n",
       "4  21.805298     1083.764450   feminino   bom"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df2 = pd.read_csv('dados/exemplo3.csv')\n",
    "df2.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>idade</th>\n",
       "      <th>conta_corrente</th>\n",
       "      <th>risco_bom</th>\n",
       "      <th>sexo_masculino</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>21.257389</td>\n",
       "      <td>783.127911</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>21.766573</td>\n",
       "      <td>979.747521</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>25.321033</td>\n",
       "      <td>1065.328054</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>23.919268</td>\n",
       "      <td>1195.758078</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>21.805298</td>\n",
       "      <td>1083.764450</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       idade  conta_corrente  risco_bom  sexo_masculino\n",
       "0  21.257389      783.127911        0.0             1.0\n",
       "1  21.766573      979.747521        1.0             0.0\n",
       "2  25.321033     1065.328054        1.0             0.0\n",
       "3  23.919268     1195.758078        1.0             0.0\n",
       "4  21.805298     1083.764450        1.0             0.0"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df2['risco_bom'] = df2.risco.apply(encoder, args=('bom',))\n",
    "df2['sexo_masculino'] = df2.sexo.apply(encoder, args=('masculino',))\n",
    "df2.drop('sexo', axis=1, inplace=True)\n",
    "df2.drop('risco', axis=1, inplace=True)\n",
    "df2.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "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>idade</th>\n",
       "      <th>conta_corrente</th>\n",
       "      <th>risco_bom</th>\n",
       "      <th>sexo_masculino</th>\n",
       "      <th>km_3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>21.257389</td>\n",
       "      <td>783.127911</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>21.766573</td>\n",
       "      <td>979.747521</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>25.321033</td>\n",
       "      <td>1065.328054</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>23.919268</td>\n",
       "      <td>1195.758078</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>21.805298</td>\n",
       "      <td>1083.764450</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       idade  conta_corrente  risco_bom  sexo_masculino  km_3\n",
       "0  21.257389      783.127911        0.0             1.0     0\n",
       "1  21.766573      979.747521        1.0             0.0     1\n",
       "2  25.321033     1065.328054        1.0             0.0     1\n",
       "3  23.919268     1195.758078        1.0             0.0     1\n",
       "4  21.805298     1083.764450        1.0             0.0     1"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = MinMaxScaler().fit_transform(df2[['idade', 'conta_corrente', 'risco_bom', 'sexo_masculino']])\n",
    "df2['km_3'] = KMeans(n_clusters=3, random_state=1).fit_predict(X)\n",
    "df2.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cluster 0\n",
      "- idade: mean = 33.06556739891999\n",
      "- conta_corrente: mean = 2099.935581556636\n",
      "- risco_bom: count = 0.0 %\n",
      "- sexo_masculino: count = 59.04761904761905 %\n",
      "Cluster 1\n",
      "- idade: mean = 31.13198723733135\n",
      "- conta_corrente: mean = 3043.8601326847474\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 0.0 %\n",
      "Cluster 2\n",
      "- idade: mean = 30.934583754682933\n",
      "- conta_corrente: mean = 3301.585452141532\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 100.0 %\n"
     ]
    }
   ],
   "source": [
    "cluster_report(df2, df2.columns.values[0:4], 'km_3')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "df2['km_4'] = KMeans(n_clusters=4, random_state=1).fit_predict(X)\n",
    "df2['km_5'] = KMeans(n_clusters=5, random_state=1).fit_predict(X)\n",
    "df2['km_6'] = KMeans(n_clusters=6, random_state=1).fit_predict(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cluster 0\n",
      "- idade: mean = 23.04249450448745\n",
      "- conta_corrente: mean = 885.6266167706968\n",
      "- risco_bom: count = 0.0 %\n",
      "- sexo_masculino: count = 0.0 %\n",
      "Cluster 1\n",
      "- idade: mean = 32.854090296333226\n",
      "- conta_corrente: mean = 5120.53892024885\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 0.0 %\n",
      "Cluster 2\n",
      "- idade: mean = 32.41618857720974\n",
      "- conta_corrente: mean = 5141.148601937596\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 100.0 %\n",
      "Cluster 3\n",
      "- idade: mean = 23.454019079971687\n",
      "- conta_corrente: mean = 902.733633147001\n",
      "- risco_bom: count = 0.0 %\n",
      "- sexo_masculino: count = 100.0 %\n",
      "Cluster 4\n",
      "- idade: mean = 22.162928520248137\n",
      "- conta_corrente: mean = 1127.7377476345093\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 100.0 %\n",
      "Cluster 5\n",
      "- idade: mean = 22.66954579082227\n",
      "- conta_corrente: mean = 1044.6261074209237\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 0.0 %\n",
      "Cluster 6\n",
      "- idade: mean = 46.132040576999614\n",
      "- conta_corrente: mean = 3617.3431555573716\n",
      "- risco_bom: count = 0.0 %\n",
      "- sexo_masculino: count = 100.0 %\n",
      "Cluster 7\n",
      "- idade: mean = 44.974433434322584\n",
      "- conta_corrente: mean = 3701.6971261042777\n",
      "- risco_bom: count = 0.0 %\n",
      "- sexo_masculino: count = 0.0 %\n",
      "Cluster 8\n",
      "- idade: mean = 55.23334953209528\n",
      "- conta_corrente: mean = 3119.618058167428\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 0.0 %\n",
      "Cluster 9\n",
      "- idade: mean = 57.37081185675326\n",
      "- conta_corrente: mean = 3294.386412232267\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 100.0 %\n"
     ]
    }
   ],
   "source": [
    "k = 10\n",
    "df2[f'km_{k}'] = KMeans(n_clusters=k, random_state=1).fit_predict(X)\n",
    "cluster_report(df2, df2.columns.values[0:4], f'km_{k}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cluster_report_perc(df, columns, cluster_column):\n",
    "    for c in sorted(df[cluster_column].unique()):\n",
    "        print(f'Cluster {c}')\n",
    "        for column in columns:\n",
    "            if len(df[column].unique()) > 2:\n",
    "                value = df[df[cluster_column] == c][column].mean()\n",
    "                ref = df[column].mean()\n",
    "                print(f'- {column}: mean = {round((value - ref) / ref * 100, 2)} %')\n",
    "            else:\n",
    "                n = df[df[cluster_column] == c][column].mean()\n",
    "                print(f'- {column}: count = {n * 100} %')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cluster 0\n",
      "- idade: mean = 19.33 %\n",
      "- conta_corrente: mean = 69.12 %\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 0.0 %\n",
      "Cluster 1\n",
      "- idade: mean = 5.3 %\n",
      "- conta_corrente: mean = -23.76 %\n",
      "- risco_bom: count = 0.0 %\n",
      "- sexo_masculino: count = 100.0 %\n",
      "Cluster 2\n",
      "- idade: mean = 16.3 %\n",
      "- conta_corrente: mean = 73.99 %\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 100.0 %\n",
      "Cluster 3\n",
      "- idade: mean = 2.02 %\n",
      "- conta_corrente: mean = -24.15 %\n",
      "- risco_bom: count = 0.0 %\n",
      "- sexo_masculino: count = 0.0 %\n",
      "Cluster 4\n",
      "- idade: mean = -28.5 %\n",
      "- conta_corrente: mean = -62.08 %\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 0.0 %\n",
      "Cluster 5\n",
      "- idade: mean = -30.32 %\n",
      "- conta_corrente: mean = -59.14 %\n",
      "- risco_bom: count = 100.0 %\n",
      "- sexo_masculino: count = 100.0 %\n"
     ]
    }
   ],
   "source": [
    "cluster_report_perc(df2, df2.columns.values[0:4], 'km_6')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cluster_report_perc_df(df, columns, cluster_column):\n",
    "    data = {}\n",
    "    for c in sorted(df[cluster_column].unique()):\n",
    "        data[f'cluster {c}'] = []\n",
    "        for column in columns:\n",
    "            if len(df[column].unique()) > 2:\n",
    "                value = df[df[cluster_column] == c][column].mean()\n",
    "                ref = df[column].mean()\n",
    "                data[f'cluster {c}'].append(round((value - ref) / ref * 100, 2))\n",
    "            else:\n",
    "                n = df[df[cluster_column] == c][column].mean()\n",
    "                data[f'cluster {c}'].append(n * 100)\n",
    "    return pd.DataFrame(data, index=columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "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>idade</th>\n",
       "      <th>conta_corrente</th>\n",
       "      <th>risco_bom</th>\n",
       "      <th>sexo_masculino</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>cluster 0</td>\n",
       "      <td>19.33</td>\n",
       "      <td>69.12</td>\n",
       "      <td>100.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>cluster 1</td>\n",
       "      <td>5.30</td>\n",
       "      <td>-23.76</td>\n",
       "      <td>0.0</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>cluster 2</td>\n",
       "      <td>16.30</td>\n",
       "      <td>73.99</td>\n",
       "      <td>100.0</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>cluster 3</td>\n",
       "      <td>2.02</td>\n",
       "      <td>-24.15</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>cluster 4</td>\n",
       "      <td>-28.50</td>\n",
       "      <td>-62.08</td>\n",
       "      <td>100.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>cluster 5</td>\n",
       "      <td>-30.32</td>\n",
       "      <td>-59.14</td>\n",
       "      <td>100.0</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           idade  conta_corrente  risco_bom  sexo_masculino\n",
       "cluster 0  19.33           69.12      100.0             0.0\n",
       "cluster 1   5.30          -23.76        0.0           100.0\n",
       "cluster 2  16.30           73.99      100.0           100.0\n",
       "cluster 3   2.02          -24.15        0.0             0.0\n",
       "cluster 4 -28.50          -62.08      100.0             0.0\n",
       "cluster 5 -30.32          -59.14      100.0           100.0"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results = cluster_report_perc_df(df2, df2.columns.values[0:4], 'km_6')\n",
    "results = results.transpose()\n",
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import seaborn as sns\n",
    "\n",
    "plt.figure(figsize=(15, 6))\n",
    "s = sns.heatmap(results, linewidths=.5)"
   ]
  }
 ],
 "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}