notebook/2_Model_Training_Gemstone.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">2. Model Training for Gemstone Dataset - Utkarsh Gaikwad</p>"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Loading the necessary Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Basic Import\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt \n",
    "import seaborn as sns\n",
    "# Modelling\n",
    "from sklearn.neighbors import KNeighborsRegressor\n",
    "from sklearn.tree import DecisionTreeRegressor\n",
    "from sklearn.ensemble import RandomForestRegressor,AdaBoostRegressor\n",
    "from sklearn.svm import SVR\n",
    "from sklearn.linear_model import LinearRegression, Ridge,Lasso\n",
    "from sklearn.metrics import r2_score, mean_absolute_error, mean_squared_error\n",
    "from sklearn.model_selection import RandomizedSearchCV\n",
    "from catboost import CatBoostRegressor\n",
    "from xgboost import XGBRegressor"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">Read the Dataset</p>"
   ]
  },
  {
   "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>id</th>\n",
       "      <th>carat</th>\n",
       "      <th>cut</th>\n",
       "      <th>color</th>\n",
       "      <th>clarity</th>\n",
       "      <th>depth</th>\n",
       "      <th>table</th>\n",
       "      <th>x</th>\n",
       "      <th>y</th>\n",
       "      <th>z</th>\n",
       "      <th>price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1.52</td>\n",
       "      <td>Premium</td>\n",
       "      <td>F</td>\n",
       "      <td>VS2</td>\n",
       "      <td>62.2</td>\n",
       "      <td>58.0</td>\n",
       "      <td>7.27</td>\n",
       "      <td>7.33</td>\n",
       "      <td>4.55</td>\n",
       "      <td>13619</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2.03</td>\n",
       "      <td>Very Good</td>\n",
       "      <td>J</td>\n",
       "      <td>SI2</td>\n",
       "      <td>62.0</td>\n",
       "      <td>58.0</td>\n",
       "      <td>8.06</td>\n",
       "      <td>8.12</td>\n",
       "      <td>5.05</td>\n",
       "      <td>13387</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>0.70</td>\n",
       "      <td>Ideal</td>\n",
       "      <td>G</td>\n",
       "      <td>VS1</td>\n",
       "      <td>61.2</td>\n",
       "      <td>57.0</td>\n",
       "      <td>5.69</td>\n",
       "      <td>5.73</td>\n",
       "      <td>3.50</td>\n",
       "      <td>2772</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>0.32</td>\n",
       "      <td>Ideal</td>\n",
       "      <td>G</td>\n",
       "      <td>VS1</td>\n",
       "      <td>61.6</td>\n",
       "      <td>56.0</td>\n",
       "      <td>4.38</td>\n",
       "      <td>4.41</td>\n",
       "      <td>2.71</td>\n",
       "      <td>666</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>1.70</td>\n",
       "      <td>Premium</td>\n",
       "      <td>G</td>\n",
       "      <td>VS2</td>\n",
       "      <td>62.6</td>\n",
       "      <td>59.0</td>\n",
       "      <td>7.65</td>\n",
       "      <td>7.61</td>\n",
       "      <td>4.77</td>\n",
       "      <td>14453</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   id  carat        cut color clarity  depth  table     x     y     z  price\n",
       "0   0   1.52    Premium     F     VS2   62.2   58.0  7.27  7.33  4.55  13619\n",
       "1   1   2.03  Very Good     J     SI2   62.0   58.0  8.06  8.12  5.05  13387\n",
       "2   2   0.70      Ideal     G     VS1   61.2   57.0  5.69  5.73  3.50   2772\n",
       "3   3   0.32      Ideal     G     VS1   61.6   56.0  4.38  4.41  2.71    666\n",
       "4   4   1.70    Premium     G     VS2   62.6   59.0  7.65  7.61  4.77  14453"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv('./data/gemstone.csv')\n",
    "df.head()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dropping id Column"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.drop(labels=['id'],axis=1)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">Getting X and Y variables</p>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df.drop(labels=['price'],axis=1)\n",
    "Y = df[['price']]"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">Creating Data Transformation Pipeline</p>"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Creating Pipeline with Column Transformer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define which columns should be ordinal-encoded and which should be scaled\n",
    "categorical_cols = X.select_dtypes(include='object').columns\n",
    "numerical_cols = X.select_dtypes(exclude='object').columns\n",
    "            \n",
    "# Define the custom ranking for each ordinal variable\n",
    "cut_categories = ['Fair', 'Good', 'Very Good','Premium','Ideal']\n",
    "color_categories = ['D', 'E', 'F', 'G', 'H', 'I', 'J']\n",
    "clarity_categories = ['I1','SI2','SI1','VS2','VS1','VVS2','VVS1','IF']\n",
    "\n",
    "from sklearn.compose import ColumnTransformer\n",
    "from sklearn.impute import SimpleImputer\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.preprocessing import OrdinalEncoder,StandardScaler\n",
    "\n",
    "# Numerical Pipeline\n",
    "num_pipeline = Pipeline(\n",
    "                steps = [\n",
    "                ('imputer',SimpleImputer(strategy='median')),\n",
    "                ('scaler',StandardScaler())                \n",
    "                ]\n",
    "            )\n",
    "\n",
    "# Categorical Pipeline\n",
    "cat_pipeline = Pipeline(\n",
    "                steps=[\n",
    "                ('imputer',SimpleImputer(strategy='most_frequent')),\n",
    "                ('ordinal_encoder',OrdinalEncoder(categories=[cut_categories,color_categories,clarity_categories])),\n",
    "                ('scaler',StandardScaler())\n",
    "                ]\n",
    "            )\n",
    "\n",
    "preprocessor = ColumnTransformer(\n",
    "                [\n",
    "                ('num_pipeline',num_pipeline,numerical_cols),\n",
    "                ('cat_pipeline',cat_pipeline,categorical_cols)\n",
    "                ]\n",
    "            )"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">Train Test Split</p>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "xtrain, xtest, ytrain, ytest = train_test_split(X,Y,test_size=0.2,random_state=42)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">Transforming the data with pipeline created</p>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "xtrain = pd.DataFrame(preprocessor.fit_transform(xtrain),columns=preprocessor.get_feature_names_out())\n",
    "xtest = pd.DataFrame(preprocessor.transform(xtest),columns=preprocessor.get_feature_names_out())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['num_pipeline__carat', 'num_pipeline__depth',\n",
       "       'num_pipeline__table', 'num_pipeline__x', 'num_pipeline__y',\n",
       "       'num_pipeline__z', 'cat_pipeline__cut', 'cat_pipeline__color',\n",
       "       'cat_pipeline__clarity'], dtype=object)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preprocessor.get_feature_names_out()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>num_pipeline__carat</th>\n",
       "      <th>num_pipeline__depth</th>\n",
       "      <th>num_pipeline__table</th>\n",
       "      <th>num_pipeline__x</th>\n",
       "      <th>num_pipeline__y</th>\n",
       "      <th>num_pipeline__z</th>\n",
       "      <th>cat_pipeline__cut</th>\n",
       "      <th>cat_pipeline__color</th>\n",
       "      <th>cat_pipeline__clarity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-1.016395</td>\n",
       "      <td>-0.204317</td>\n",
       "      <td>0.402608</td>\n",
       "      <td>-1.202472</td>\n",
       "      <td>-1.187395</td>\n",
       "      <td>-1.194148</td>\n",
       "      <td>-0.132842</td>\n",
       "      <td>-0.936018</td>\n",
       "      <td>-0.648950</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.882396</td>\n",
       "      <td>0.720758</td>\n",
       "      <td>-0.118536</td>\n",
       "      <td>0.985177</td>\n",
       "      <td>0.941823</td>\n",
       "      <td>1.036109</td>\n",
       "      <td>-0.132842</td>\n",
       "      <td>-0.320002</td>\n",
       "      <td>0.017052</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1.529711</td>\n",
       "      <td>0.350728</td>\n",
       "      <td>-1.160823</td>\n",
       "      <td>1.426308</td>\n",
       "      <td>1.394848</td>\n",
       "      <td>1.441611</td>\n",
       "      <td>0.872563</td>\n",
       "      <td>1.528047</td>\n",
       "      <td>0.017052</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1.896523</td>\n",
       "      <td>0.073206</td>\n",
       "      <td>0.923751</td>\n",
       "      <td>1.741402</td>\n",
       "      <td>1.711965</td>\n",
       "      <td>1.702290</td>\n",
       "      <td>-0.132842</td>\n",
       "      <td>1.528047</td>\n",
       "      <td>-1.314953</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.450852</td>\n",
       "      <td>1.738340</td>\n",
       "      <td>1.444895</td>\n",
       "      <td>0.562052</td>\n",
       "      <td>0.525040</td>\n",
       "      <td>0.703019</td>\n",
       "      <td>-2.143651</td>\n",
       "      <td>0.912031</td>\n",
       "      <td>0.017052</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   num_pipeline__carat  num_pipeline__depth  num_pipeline__table  \\\n",
       "0            -1.016395            -0.204317             0.402608   \n",
       "1             0.882396             0.720758            -0.118536   \n",
       "2             1.529711             0.350728            -1.160823   \n",
       "3             1.896523             0.073206             0.923751   \n",
       "4             0.450852             1.738340             1.444895   \n",
       "\n",
       "   num_pipeline__x  num_pipeline__y  num_pipeline__z  cat_pipeline__cut  \\\n",
       "0        -1.202472        -1.187395        -1.194148          -0.132842   \n",
       "1         0.985177         0.941823         1.036109          -0.132842   \n",
       "2         1.426308         1.394848         1.441611           0.872563   \n",
       "3         1.741402         1.711965         1.702290          -0.132842   \n",
       "4         0.562052         0.525040         0.703019          -2.143651   \n",
       "\n",
       "   cat_pipeline__color  cat_pipeline__clarity  \n",
       "0            -0.936018              -0.648950  \n",
       "1            -0.320002               0.017052  \n",
       "2             1.528047               0.017052  \n",
       "3             1.528047              -1.314953  \n",
       "4             0.912031               0.017052  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xtrain.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "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>num_pipeline__carat</th>\n",
       "      <th>num_pipeline__depth</th>\n",
       "      <th>num_pipeline__table</th>\n",
       "      <th>num_pipeline__x</th>\n",
       "      <th>num_pipeline__y</th>\n",
       "      <th>num_pipeline__z</th>\n",
       "      <th>cat_pipeline__cut</th>\n",
       "      <th>cat_pipeline__color</th>\n",
       "      <th>cat_pipeline__clarity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.628006</td>\n",
       "      <td>0.258221</td>\n",
       "      <td>-0.118536</td>\n",
       "      <td>-0.599293</td>\n",
       "      <td>-0.580341</td>\n",
       "      <td>-0.571414</td>\n",
       "      <td>0.872563</td>\n",
       "      <td>-1.552034</td>\n",
       "      <td>-0.648950</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2.608569</td>\n",
       "      <td>-2.146973</td>\n",
       "      <td>-0.118536</td>\n",
       "      <td>2.128516</td>\n",
       "      <td>2.201232</td>\n",
       "      <td>1.962969</td>\n",
       "      <td>-1.138246</td>\n",
       "      <td>0.296015</td>\n",
       "      <td>-1.314953</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-1.124281</td>\n",
       "      <td>-1.221899</td>\n",
       "      <td>0.923751</td>\n",
       "      <td>-1.373523</td>\n",
       "      <td>-1.413907</td>\n",
       "      <td>-1.469310</td>\n",
       "      <td>-0.132842</td>\n",
       "      <td>-0.936018</td>\n",
       "      <td>2.015061</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-1.016395</td>\n",
       "      <td>-0.574346</td>\n",
       "      <td>0.923751</td>\n",
       "      <td>-1.157458</td>\n",
       "      <td>-1.160213</td>\n",
       "      <td>-1.194148</td>\n",
       "      <td>-0.132842</td>\n",
       "      <td>1.528047</td>\n",
       "      <td>2.015061</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.860819</td>\n",
       "      <td>0.628251</td>\n",
       "      <td>-0.639679</td>\n",
       "      <td>0.949167</td>\n",
       "      <td>0.987125</td>\n",
       "      <td>1.007145</td>\n",
       "      <td>0.872563</td>\n",
       "      <td>0.912031</td>\n",
       "      <td>-0.648950</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   num_pipeline__carat  num_pipeline__depth  num_pipeline__table  \\\n",
       "0            -0.628006             0.258221            -0.118536   \n",
       "1             2.608569            -2.146973            -0.118536   \n",
       "2            -1.124281            -1.221899             0.923751   \n",
       "3            -1.016395            -0.574346             0.923751   \n",
       "4             0.860819             0.628251            -0.639679   \n",
       "\n",
       "   num_pipeline__x  num_pipeline__y  num_pipeline__z  cat_pipeline__cut  \\\n",
       "0        -0.599293        -0.580341        -0.571414           0.872563   \n",
       "1         2.128516         2.201232         1.962969          -1.138246   \n",
       "2        -1.373523        -1.413907        -1.469310          -0.132842   \n",
       "3        -1.157458        -1.160213        -1.194148          -0.132842   \n",
       "4         0.949167         0.987125         1.007145           0.872563   \n",
       "\n",
       "   cat_pipeline__color  cat_pipeline__clarity  \n",
       "0            -1.552034              -0.648950  \n",
       "1             0.296015              -1.314953  \n",
       "2            -0.936018               2.015061  \n",
       "3             1.528047               2.015061  \n",
       "4             0.912031              -0.648950  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xtest.head()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">Model Training Baseline models</p>"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create an Evaluate Function to give all metrics after model Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate_model(true, predicted):\n",
    "    mae = mean_absolute_error(true, predicted)\n",
    "    mse = mean_squared_error(true, predicted)\n",
    "    rmse = np.sqrt(mean_squared_error(true, predicted))\n",
    "    r2_square = r2_score(true, predicted)\n",
    "    return mae, rmse, r2_square"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training Various models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Linear Regression\n",
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 1016.9490\n",
      "- Mean Absolute Error: 677.1656\n",
      "- R2 Score: 0.9366\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 1006.6010\n",
      "- Mean Absolute Error: 671.5856\n",
      "- R2 Score: 0.9373\n",
      "===================================\n",
      "\n",
      "\n",
      "Lasso\n",
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 1017.0718\n",
      "- Mean Absolute Error: 678.3145\n",
      "- R2 Score: 0.9366\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 1006.8716\n",
      "- Mean Absolute Error: 672.8635\n",
      "- R2 Score: 0.9373\n",
      "===================================\n",
      "\n",
      "\n",
      "Ridge\n",
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 1016.9491\n",
      "- Mean Absolute Error: 677.1925\n",
      "- R2 Score: 0.9366\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 1006.6062\n",
      "- Mean Absolute Error: 671.6137\n",
      "- R2 Score: 0.9373\n",
      "===================================\n",
      "\n",
      "\n",
      "K-Neighbors Regressor\n",
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 545.6563\n",
      "- Mean Absolute Error: 285.8181\n",
      "- R2 Score: 0.9817\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 671.2898\n",
      "- Mean Absolute Error: 349.4687\n",
      "- R2 Score: 0.9721\n",
      "===================================\n",
      "\n",
      "\n",
      "Decision Tree\n",
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 12.8816\n",
      "- Mean Absolute Error: 0.8012\n",
      "- R2 Score: 1.0000\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 836.0930\n",
      "- Mean Absolute Error: 422.9325\n",
      "- R2 Score: 0.9567\n",
      "===================================\n",
      "\n",
      "\n",
      "Random Forest Regressor\n",
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 228.4535\n",
      "- Mean Absolute Error: 116.0063\n",
      "- R2 Score: 0.9968\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 605.5144\n",
      "- Mean Absolute Error: 308.8910\n",
      "- R2 Score: 0.9773\n",
      "===================================\n",
      "\n",
      "\n",
      "XGBRegressor\n",
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 508.4624\n",
      "- Mean Absolute Error: 274.4553\n",
      "- R2 Score: 0.9841\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 587.6792\n",
      "- Mean Absolute Error: 297.4995\n",
      "- R2 Score: 0.9786\n",
      "===================================\n",
      "\n",
      "\n",
      "CatBoosting Regressor\n",
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 530.6367\n",
      "- Mean Absolute Error: 282.2494\n",
      "- R2 Score: 0.9827\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 579.9574\n",
      "- Mean Absolute Error: 295.4844\n",
      "- R2 Score: 0.9792\n",
      "===================================\n",
      "\n",
      "\n",
      "AdaBoost Regressor\n",
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 1451.3669\n",
      "- Mean Absolute Error: 1051.3943\n",
      "- R2 Score: 0.8708\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 1460.8063\n",
      "- Mean Absolute Error: 1055.7695\n",
      "- R2 Score: 0.8679\n",
      "===================================\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "models = {\n",
    "    \"Linear Regression\": LinearRegression(),\n",
    "    \"Lasso\": Lasso(),\n",
    "    \"Ridge\": Ridge(),\n",
    "    \"K-Neighbors Regressor\": KNeighborsRegressor(),\n",
    "    \"Decision Tree\": DecisionTreeRegressor(),\n",
    "    \"Random Forest Regressor\": RandomForestRegressor(),\n",
    "    \"XGBRegressor\": XGBRegressor(), \n",
    "    \"CatBoosting Regressor\": CatBoostRegressor(verbose=False),\n",
    "    \"AdaBoost Regressor\": AdaBoostRegressor()\n",
    "}\n",
    "model_list = []\n",
    "r2_list =[]\n",
    "\n",
    "for i in range(len(list(models))):\n",
    "    model = list(models.values())[i]\n",
    "    model.fit(xtrain, ytrain.values.flatten()) # Train model\n",
    "\n",
    "    # Make predictions\n",
    "    y_train_pred = model.predict(xtrain)\n",
    "    y_test_pred = model.predict(xtest)\n",
    "    \n",
    "    # Evaluate Train and Test dataset\n",
    "    model_train_mae , model_train_rmse, model_train_r2 = evaluate_model(ytrain, y_train_pred)\n",
    "\n",
    "    model_test_mae , model_test_rmse, model_test_r2 = evaluate_model(ytest, y_test_pred)\n",
    "\n",
    "    \n",
    "    print(list(models.keys())[i])\n",
    "    model_list.append(list(models.keys())[i])\n",
    "    \n",
    "    print('Model performance for Training set')\n",
    "    print(\"- Root Mean Squared Error: {:.4f}\".format(model_train_rmse))\n",
    "    print(\"- Mean Absolute Error: {:.4f}\".format(model_train_mae))\n",
    "    print(\"- R2 Score: {:.4f}\".format(model_train_r2))\n",
    "\n",
    "    print('----------------------------------')\n",
    "    \n",
    "    print('Model performance for Test set')\n",
    "    print(\"- Root Mean Squared Error: {:.4f}\".format(model_test_rmse))\n",
    "    print(\"- Mean Absolute Error: {:.4f}\".format(model_test_mae))\n",
    "    print(\"- R2 Score: {:.4f}\".format(model_test_r2))\n",
    "    r2_list.append(model_test_r2)\n",
    "    \n",
    "    print('='*35)\n",
    "    print('\\n')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "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>Model Name</th>\n",
       "      <th>R2_Score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>CatBoosting Regressor</td>\n",
       "      <td>0.979186</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>XGBRegressor</td>\n",
       "      <td>0.978628</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Random Forest Regressor</td>\n",
       "      <td>0.977311</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>K-Neighbors Regressor</td>\n",
       "      <td>0.972114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Decision Tree</td>\n",
       "      <td>0.956741</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Linear Regression</td>\n",
       "      <td>0.937298</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Ridge</td>\n",
       "      <td>0.937297</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Lasso</td>\n",
       "      <td>0.937264</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>AdaBoost Regressor</td>\n",
       "      <td>0.867945</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                Model Name  R2_Score\n",
       "7    CatBoosting Regressor  0.979186\n",
       "6             XGBRegressor  0.978628\n",
       "5  Random Forest Regressor  0.977311\n",
       "3    K-Neighbors Regressor  0.972114\n",
       "4            Decision Tree  0.956741\n",
       "0        Linear Regression  0.937298\n",
       "2                    Ridge  0.937297\n",
       "1                    Lasso  0.937264\n",
       "8       AdaBoost Regressor  0.867945"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_results = pd.DataFrame(list(zip(model_list, r2_list)), columns=['Model Name', 'R2_Score']).sort_values(by=[\"R2_Score\"],ascending=False)\n",
    "df_results"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">Hyperparameter tuning</p>"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tuning Catboost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'learning_rate': 0.04, 'iterations': 600, 'depth': 7}\n",
      "0.9797327917338592\n"
     ]
    }
   ],
   "source": [
    "# Initializing catboost\n",
    "cbr = CatBoostRegressor(verbose=False)\n",
    "\n",
    "# Creating the hyperparameter grid\n",
    "param_dist = {'depth'          : [4,5,6,7,8,9, 10],\n",
    "              'learning_rate' : [0.01,0.02,0.03,0.04],\n",
    "               'iterations'    : [300,400,500,600]}\n",
    "\n",
    "#Instantiate RandomSearchCV object\n",
    "rscv = RandomizedSearchCV(cbr , param_dist, scoring='r2', cv =5, n_jobs=-1)\n",
    "\n",
    "# Fit the model\n",
    "rscv.fit(xtrain, ytrain.values.flatten())\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(rscv.best_params_)\n",
    "print(rscv.best_score_)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Definition to print evaluated model results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_evaluated_results(model,xtrain,ytrain,xtest,ytest):\n",
    "    ytrain_pred = model.predict(xtrain)\n",
    "    ytest_pred = model.predict(xtest)\n",
    "\n",
    "    # Evaluate Train and Test dataset\n",
    "    model_train_mae , model_train_rmse, model_train_r2 = evaluate_model(ytrain, ytrain_pred)\n",
    "    model_test_mae , model_test_rmse, model_test_r2 = evaluate_model(ytest, ytest_pred)\n",
    "\n",
    "    # Printing results\n",
    "    print('Model performance for Training set')\n",
    "    print(\"- Root Mean Squared Error: {:.4f}\".format(model_train_rmse))\n",
    "    print(\"- Mean Absolute Error: {:.4f}\".format(model_train_mae))\n",
    "    print(\"- R2 Score: {:.4f}\".format(model_train_r2))\n",
    "\n",
    "    print('----------------------------------')\n",
    "    \n",
    "    print('Model performance for Test set')\n",
    "    print(\"- Root Mean Squared Error: {:.4f}\".format(model_test_rmse))\n",
    "    print(\"- Mean Absolute Error: {:.4f}\".format(model_test_mae))\n",
    "    print(\"- R2 Score: {:.4f}\".format(model_test_r2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 555.7681\n",
      "- Mean Absolute Error: 293.2188\n",
      "- R2 Score: 0.9811\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 576.7162\n",
      "- Mean Absolute Error: 297.2402\n",
      "- R2 Score: 0.9794\n"
     ]
    }
   ],
   "source": [
    "# Selecting best model\n",
    "best_cbr = rscv.best_estimator_\n",
    "\n",
    "# Evaluate Train and Test dataset\n",
    "print_evaluated_results(best_cbr,xtrain,ytrain,xtest,ytest)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tuning KNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'n_neighbors': 16}\n",
      "0.9736029547852756\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "# Initialize knn\n",
    "knn = KNeighborsRegressor()\n",
    "\n",
    "# parameters\n",
    "k_range = list(range(2, 31))\n",
    "param_grid = dict(n_neighbors=k_range)\n",
    "\n",
    "# Fitting the cvmodel\n",
    "grid = GridSearchCV(knn, param_grid, cv=5, scoring='r2',n_jobs=-1)\n",
    "grid.fit(xtrain, ytrain)\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(grid.best_params_)\n",
    "print(grid.best_score_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 610.3458\n",
      "- Mean Absolute Error: 319.3932\n",
      "- R2 Score: 0.9772\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 643.8459\n",
      "- Mean Absolute Error: 336.3468\n",
      "- R2 Score: 0.9743\n"
     ]
    }
   ],
   "source": [
    "# Selecting best model\n",
    "best_knn = grid.best_estimator_\n",
    "\n",
    "# Evaluate Train and Test dataset\n",
    "print_evaluated_results(best_knn,xtrain,ytrain,xtest,ytest)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tuning XGBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'n_estimators': 500, 'min_child_weight': 1, 'max_depth': 4, 'learning_rate': 0.2, 'gamma': 0.1, 'colsample_bytree': 0.4}\n",
      "0.978786564716807\n"
     ]
    }
   ],
   "source": [
    "# Initializing xgboost\n",
    "xgb = XGBRegressor()\n",
    "\n",
    "# Parameters\n",
    "params = {\n",
    " 'learning_rate' : [0.05,0.10,0.15,0.20,0.25,0.30],\n",
    " 'max_depth' : [ 3, 4, 5, 6, 8, 10, 12, 15],\n",
    " 'min_child_weight' : [ 1, 3, 5, 7 ],\n",
    " 'gamma': [ 0.0, 0.1, 0.2 , 0.3, 0.4 ],\n",
    " 'colsample_bytree' : [ 0.3, 0.4, 0.5 , 0.7 ],\n",
    " 'n_estimators':[300,400,500,600]\n",
    "}\n",
    "\n",
    "rs_xgb=RandomizedSearchCV(xgb,param_distributions=params,scoring='r2',n_jobs=-1,cv=5)\n",
    "rs_xgb.fit(xtrain, ytrain.values.flatten())\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(rs_xgb.best_params_)\n",
    "print(rs_xgb.best_score_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 540.2572\n",
      "- Mean Absolute Error: 288.3935\n",
      "- R2 Score: 0.9821\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 587.9065\n",
      "- Mean Absolute Error: 301.7455\n",
      "- R2 Score: 0.9786\n"
     ]
    }
   ],
   "source": [
    "# Selecting best xgb model\n",
    "best_xgb = rs_xgb.best_estimator_\n",
    "\n",
    "# Evaluate Train and Test dataset\n",
    "print_evaluated_results(best_xgb,xtrain,ytrain,xtest,ytest)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">Voting Regressor</p>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model performance for Training set\n",
      "- Root Mean Squared Error: 532.5211\n",
      "- Mean Absolute Error: 281.9885\n",
      "- R2 Score: 0.9826\n",
      "----------------------------------\n",
      "Model performance for Test set\n",
      "- Root Mean Squared Error: 575.7354\n",
      "- Mean Absolute Error: 293.7040\n",
      "- R2 Score: 0.9795\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import VotingRegressor\n",
    "\n",
    "er = VotingRegressor([('cbr',best_cbr),('knn',best_knn),('xgb',XGBRegressor())], weights=[3,1,2])\n",
    "er.fit(xtrain, ytrain.values.flatten())\n",
    "\n",
    "print_evaluated_results(er,xtrain,ytrain,xtest,ytest)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model improved with Ensemble technique !"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <p style=\"padding:10px;background-color:#87CEEB ;margin:10;color:#000000;font-family:newtimeroman;font-size:100%;text-align:center;border-radius: 10px 10px ;overflow:hidden;font-weight:50\">Catboost Model Feature Importances</p>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "num_pipeline__y          21.720701\n",
      "num_pipeline__carat      19.408904\n",
      "num_pipeline__x          18.456375\n",
      "num_pipeline__z          16.001919\n",
      "cat_pipeline__clarity    14.287774\n",
      "cat_pipeline__color       9.674735\n",
      "cat_pipeline__cut         0.247507\n",
      "num_pipeline__table       0.103163\n",
      "num_pipeline__depth       0.098922\n",
      "dtype: float64\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAArQAAAHHCAYAAABOYMsoAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABwyElEQVR4nO3deVwV1f8/8NeIcL3sISCgIBKIuKAIKmgqLoVmhmm5ROGCu4S48vGjBJiF2QfFJUkzRUvB5aOWkZmZmLuWQh+VUAlECyU3kJRFOL8//DFfryzeqxfxwuv5eMzjwcycOec9M/fK28OZM5IQQoCIiIiISEc1qO0AiIiIiIieBhNaIiIiItJpTGiJiIiISKcxoSUiIiIincaEloiIiIh0GhNaIiIiItJpTGiJiIiISKcxoSUiIiIincaEloiIiIh0GhNaIqKnFB8fD0mS8Msvv9R2KFQDJElCZGRkbYdBRNVgQktEdVZGRgYmTJgAJycnNGrUCKampujWrRuWLl2Ke/fuaVzfypUrER8fr/1Aa5imcUuSVOliY2NTI/HdvXsXkZGRSE5OrpH667tNmzYhNja2tsMgqlENazsAIqKakJSUhLfeegsKhQKBgYFo27YtiouLcejQIcyaNQtnz57F6tWrNapz5cqVsLS0xKhRo2om6BryJHG//PLLCAwMVNmmVCq1HNkDd+/eRVRUFADA19e3Rtp4Gvfu3UPDhrr763LTpk04c+YMQkNDazsUohqju99QIqIqZGZmYvjw4WjevDl++ukn2NrayvumTJmCixcvIikpqRYjfP61bNkS77zzTm2H8VTu37+PsrIyGBgYPFU9jRo10lJEz9Y///wDIyOj2g6D6JngkAMiqnMWLVqEgoICfPHFFyrJbDlnZ2dMnTpVXl+3bh169+4Na2trKBQKtG7dGnFxcSrHODo64uzZszhw4ID8J/hHexPv3r2LCRMmoHHjxjA1NUVgYCBu3bpVof2VK1eiTZs2UCgUsLOzw5QpU3D79u0K5bZu3QpPT08olUpYWlrinXfewZ9//qlS5urVqxg9ejSaNWsGhUIBW1tb+Pv7IysrS+24n8Sff/6JMWPGoEmTJlAoFGjTpg3Wrl2rUqa4uBjvv/8+PD09YWZmBiMjI3Tv3h379++Xy2RlZcHKygoAEBUVJcdYPmbV19e30nhHjRoFR0dHlXokScJ//vMfxMbG4sUXX4RCocC5c+cAAL///jvefPNNWFhYoFGjRvDy8sI333yj1rk+OoY2MjISkiTh/PnzeOedd2BmZgYrKyuEh4dDCIHLly/D398fpqamsLGxQUxMjEp9ycnJkCQJmzdvxr///W/Y2NjAyMgIr7/+Oi5fvlyhfXU+B6NGjYKxsTEyMjLw6quvwsTEBAEBAfD19UVSUhIuXbokX9vy66bO/Xn02q5evVq+tp06dcLJkycrxPv7779j6NChsLKyglKphKurK+bOnatSRp3PDwAsX74cbdq0gaGhIV544QV4eXlh06ZN1d4vqp/YQ0tEdc6uXbvg5OSErl27qlU+Li4Obdq0weuvv46GDRti165dmDx5MsrKyjBlyhQAQGxsLN577z0YGxvLv5ybNGmiUk9wcDDMzc0RGRmJ9PR0xMXF4dKlS3ICAzxIhqKiotC3b19MmjRJLnfy5EkcPnwY+vr6AB48aDZ69Gh06tQJ0dHRuHbtGpYuXYrDhw/j9OnTMDc3BwAMGTIEZ8+exXvvvQdHR0fk5uZi7969yM7OhqOjo1pxV6awsBDXr19X2WZiYgKFQoFr167B29sbkiQhODgYVlZW2L17N4KCgpCfny//aTs/Px9r1qzBiBEjMG7cONy5cwdffPEF/Pz8cOLECXTo0AFWVlaIi4vDpEmT8MYbb2Dw4MEAAHd3d7Xu3aPWrVuHwsJCjB8/HgqFAhYWFjh79iy6deuGpk2b4l//+heMjIywZcsWDBo0CP/973/xxhtvPFFbw4YNg5ubGxYuXIikpCQsWLAAFhYWWLVqFXr37o2PP/4YGzduxMyZM9GpUyf06NFD5fgPP/wQkiQhLCwMubm5iI2NRd++fZGSkiIP71D3cwA86JH28/PDSy+9hP/85z8wNDSEjY0N8vLycOXKFSxZsgQAYGxsDEC9+/OwTZs24c6dO5gwYQIkScKiRYswePBg/PHHH/Ln9rfffkP37t2hr6+P8ePHw9HRERkZGdi1axc+/PBDAFD78/P5558jJCQEb775JqZOnYrCwkL89ttvOH78ON5+++0numdUhwkiojokLy9PABD+/v5qH3P37t0K2/z8/ISTk5PKtjZt2oiePXtWKLtu3ToBQHh6eori4mJ5+6JFiwQA8fXXXwshhMjNzRUGBgbilVdeEaWlpXK5FStWCABi7dq1QgghiouLhbW1tWjbtq24d++eXO7bb78VAMT7778vhBDi1q1bAoD45JNPqj2/quKuCoBKl3Xr1gkhhAgKChK2trbi+vXrKscNHz5cmJmZydfz/v37oqioSKXMrVu3RJMmTcSYMWPkbX///bcAICIiIirE0rNnz0pjHzlypGjevLm8npmZKQAIU1NTkZubq1K2T58+ol27dqKwsFDeVlZWJrp27SpcXFzUuh4PxxYRESEAiPHjx8vb7t+/L5o1ayYkSRILFy5UOV+lUilGjhwpb9u/f78AIJo2bSry8/Pl7Vu2bBEAxNKlS4UQ6n8Oyq8HAPGvf/2rQvwDBgxQuVYPx6zO/Sm/to0bNxY3b96Ut3/99dcCgNi1a5e8rUePHsLExERcunRJpd6ysjL5Z3U/P/7+/qJNmzYV4iaqDIccEFGdkp+fD+BBb6K6Hn7YKS8vD9evX0fPnj3xxx9/IC8vT+16xo8fL/dUAcCkSZPQsGFDfPfddwCAH3/8EcXFxQgNDUWDBv/3z++4ceNgamoqj+v95ZdfkJubi8mTJ6uM3xwwYABatWoll1MqlTAwMEBycnKlQxuehr+/P/bu3auy+Pn5QQiB//73vxg4cCCEELh+/bq8+Pn5IS8vD6dOnQIA6OnpyeNXy8rKcPPmTdy/fx9eXl5yGW0bMmSIPIQBAG7evImffvoJQ4cOxZ07d+RYb9y4AT8/P1y4cKHCn+/VNXbsWPlnPT09eHl5QQiBoKAgebu5uTlcXV3xxx9/VDg+MDBQ5XP65ptvwtbWVv68qPs5eNikSZPUjl/T+zNs2DC88MIL8nr37t0BQD63v//+Gz///DPGjBkDBwcHlWPL/0KhyefH3NwcV65cqXRYA9GjOOSAiOoUU1NTAMCdO3fUPubw4cOIiIjA0aNHcffuXZV9eXl5MDMzU6seFxcXlXVjY2PY2trK41kvXboEAHB1dVUpZ2BgACcnJ3l/VeUAoFWrVjh06BAAQKFQ4OOPP8aMGTPQpEkTeHt747XXXkNgYOBTT7HVrFkz9O3bt8L23Nxc3L59G6tXr65ylojc3Fz55/Xr1yMmJga///47SkpK5O0tWrR4qviq8mi9Fy9ehBAC4eHhCA8PrzLepk2batzWo0mbmZkZGjVqBEtLywrbb9y4UeH4Rz8vkiTB2dn5sZ8XQPVzUK5hw4Zo1qyZRuegyf159HzLk9vy/0yVJ7Zt27atsr2///5b7c9PWFgYfvzxR3Tu3BnOzs545ZVX8Pbbb6Nbt24anCHVF0xoiahOMTU1hZ2dHc6cOaNW+YyMDPTp0wetWrXC4sWLYW9vDwMDA3z33XdYsmQJysrKajjipxMaGoqBAwdi586d2LNnD8LDwxEdHY2ffvoJHh4eWm+v/Hq88847GDlyZKVlyse/fvXVVxg1ahQGDRqEWbNmwdraGnp6eoiOjkZGRoZa7UmSBCFEhe2lpaWVln90arHyeGfOnAk/P79Kj3F2dlYrlkfp6emptQ1ApeegbQqFQqXn/3E0vT/aODdNPj9ubm5IT0/Ht99+i++//x7//e9/sXLlSrz//vvyNG9E5ZjQElGd89prr2H16tU4evQofHx8qi27a9cuFBUV4ZtvvlHpgXr0SW/g//5sWpULFy6gV69e8npBQQFycnLw6quvAgCaN28OAEhPT4eTk5Ncrri4GJmZmXKP6MPlevfurdJGenq6vL/ciy++iBkzZmDGjBm4cOECOnTogJiYGHz11Vdqxa0JKysrmJiYoLS0tNIe3Idt27YNTk5O2L59u0oMERERKuWqi++FF16o9M/15b2Xj1N+nfX19R8b77N24cIFlXUhBC5evCgndJp+DqpS1fVV9/6oq/xaV/efSU0+PwBgZGSEYcOGYdiwYSguLsbgwYPx4YcfYs6cOTo7nRrVDI6hJaI6Z/bs2TAyMsLYsWNx7dq1CvszMjKwdOlSAP/X6/RwL1NeXh7WrVtX4TgjI6NKp9cqt3r1apU/28bFxeH+/fvo378/AKBv374wMDDAsmXLVNr74osvkJeXhwEDBgAAvLy8YG1tjc8++wxFRUVyud27dyMtLU0ud/fuXRQWFqrE8OKLL8LExETluMfFrQk9PT0MGTIE//3vfytNXP7++2+VsoDqtT1+/DiOHj2qcoyhoSEAVBrjiy++iN9//12l3tTUVBw+fFiteK2treHr64tVq1YhJyen2niftQ0bNqgMjdm2bRtycnLkz4u6n4PHMTIyqnQsuLr3R11WVlbo0aMH1q5di+zsbJV95W1o8vl5dJiGgYEBWrduDSGEyveMCGAPLRHVQS+++CI2bdokT6v08JvCjhw5gq1bt8pvzXrllVdgYGCAgQMHYsKECSgoKMDnn38Oa2vrCgmQp6cn4uLisGDBAjg7O8Pa2lql56y4uBh9+vTB0KFDkZ6ejpUrV+Kll17C66+/DuDBL/w5c+YgKioK/fr1w+uvvy6X69Spk/wiA319fXz88ccYPXo0evbsiREjRsjTNTk6OmLatGkAgPPnz8vttW7dGg0bNsSOHTtw7do1DB8+XO24NbVw4ULs378fXbp0wbhx49C6dWvcvHkTp06dwo8//oibN28CeNBTvn37drzxxhsYMGAAMjMz8dlnn6F169YoKCiQ61MqlWjdujU2b96Mli1bwsLCAm3btkXbtm0xZswYLF68GH5+fggKCkJubi4+++wztGnTRn4A8HE+/fRTvPTSS2jXrh3GjRsHJycnXLt2DUePHsWVK1eQmpr6xNfiaVhYWOCll17C6NGjce3aNcTGxsLZ2Rnjxo0DoP7n4HE8PT2xefNmTJ8+HZ06dYKxsTEGDhyo9v3RxLJly/DSSy+hY8eOGD9+PFq0aIGsrCwkJSUhJSUFgPqfn1deeQU2Njbo1q0bmjRpgrS0NKxYsQIDBgzQ6KFPqiee/cQKRETPxvnz58W4ceOEo6OjMDAwECYmJqJbt25i+fLlKlM4ffPNN8Ld3V00atRIODo6io8//lisXbtWABCZmZlyuatXr4oBAwYIExMTAUCeTqp82q4DBw6I8ePHixdeeEEYGxuLgIAAcePGjQpxrVixQrRq1Uro6+uLJk2aiEmTJolbt25VKLd582bh4eEhFAqFsLCwEAEBAeLKlSvy/uvXr4spU6aIVq1aCSMjI2FmZia6dOkitmzZolJPVXFXBYCYMmVKtWWuXbsmpkyZIuzt7YW+vr6wsbERffr0EatXr5bLlJWViY8++kg0b95cKBQK4eHhIb799tsKU24JIcSRI0eEp6enMDAwqDBN1ldffSWcnJyEgYGB6NChg9izZ0+V03ZVNYVZRkaGCAwMFDY2NkJfX180bdpUvPbaa2Lbtm3Vnmf59ahs2q6///5bpdzIkSOFkZFRheN79uypMv1U+bRdCQkJYs6cOcLa2loolUoxYMCACtNdCfH4z0F1bQshREFBgXj77beFubm5ACBfN3XvT3XX9tFrI4QQZ86cEW+88YYwNzcXjRo1Eq6uriI8PFyljDqfn1WrVokePXqIxo0bC4VCIV588UUxa9YskZeXV+l5Uv0mCfEMRqoTERERgAdvCuvVqxe2bt2KN998s7bDIaoTOIaWiIiIiHQaE1oiIiIi0mlMaImIiIhIp3EMLRERERHpNPbQEhEREZFOY0JLRERERDqNL1ageqGsrAx//fUXTExMtPoaUCIiIqo5QgjcuXMHdnZ2aNCg6n5YJrRUL/z111+wt7ev7TCIiIjoCVy+fBnNmjWrcj8TWqoXyl+TePnyZZiamtZyNERERKSO/Px82NvbP/Z1x0xoqV4oH2ZgamrKhJaIiEjHPG64IB8KIyIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mmch5bqlbYRe9BAYVjbYRAREdUZWQsH1HYI7KElIiIiIt3GhJaIiIiIdBoTWiIiIiLSaUxoiYiIiEinMaElIiIiIp3GhFYHOTo6IjY2Vqt1jho1CoMGDZLXfX19ERoaqtU2iIiIiGoCp+3SQSdPnoSRkVGNtrF9+3bo6+vXaBtERERE2sCEVgdZWVnVeBsWFhY13gYRERGRNtTrIQe+vr4ICQnB7NmzYWFhARsbG0RGRgIAsrKyIEkSUlJS5PK3b9+GJElITk4GACQnJ0OSJOzZswceHh5QKpXo3bs3cnNzsXv3bri5ucHU1BRvv/027t69q3ZMwcHBCA4OhpmZGSwtLREeHg4hhFzm0SEHkiQhLi4O/fv3h1KphJOTE7Zt26ZS7+XLlzF06FCYm5vDwsIC/v7+yMrKqjaOh4ccODo64qOPPsKYMWNgYmICBwcHrF69+qnaUMfPP/8MfX19XL16VWV7aGgounfv/lR1ExERUd1QrxNaAFi/fj2MjIxw/PhxLFq0CPPnz8fevXs1qiMyMhIrVqzAkSNH5KQuNjYWmzZtQlJSEn744QcsX75co5gaNmyIEydOYOnSpVi8eDHWrFlT7THh4eEYMmQIUlNTERAQgOHDhyMtLQ0AUFJSAj8/P5iYmODgwYM4fPgwjI2N0a9fPxQXF6sdV0xMDLy8vHD69GlMnjwZkyZNQnp6ulbbeFSPHj3g5OSEL7/8Ut5WUlKCjRs3YsyYMVUeV1RUhPz8fJWFiIiI6qZ6n9C6u7sjIiICLi4uCAwMhJeXF/bt26dRHQsWLEC3bt3g4eGBoKAgHDhwAHFxcfDw8ED37t3x5ptvYv/+/WrXZ29vjyVLlsDV1RUBAQF47733sGTJkmqPeeuttzB27Fi0bNkSH3zwAby8vOQkevPmzSgrK8OaNWvQrl07uLm5Yd26dcjOzpZ7m9Xx6quvYvLkyXB2dkZYWBgsLS3l89JWG5UJCgrCunXr5PVdu3ahsLAQQ4cOrfKY6OhomJmZyYu9vf1TxUBERETPLya07u4q67a2tsjNzX3iOpo0aQJDQ0M4OTmpbNOkTm9vb0iSJK/7+PjgwoULKC0trfIYHx+fCuvlPbSpqam4ePEiTExMYGxsDGNjY1hYWKCwsBAZGRlqx/XweUqSBBsbG/m8tNVGZUaNGoWLFy/i2LFjAID4+HgMHTq02gfj5syZg7y8PHm5fPnyU8VAREREz696/1DYo0/yS5KEsrIyNGjwINd/eOxqSUnJY+uQJKnKOmtLQUEBPD09sXHjxgr7NHnArLrz0lYblbG2tsbAgQOxbt06tGjRArt3735sr69CoYBCoXiqdomIiEg31PuEtirlSVhOTg48PDwAQOUBsZp0/PhxlfVjx47BxcUFenp6VR5z7NgxBAYGqqyXx92xY0ds3rwZ1tbWMDU1rZGYa7qNsWPHYsSIEWjWrBlefPFFdOvWTettEBERkW6q90MOqqJUKuHt7Y2FCxciLS0NBw4cwLx5855J29nZ2Zg+fTrS09ORkJCA5cuXY+rUqdUes3XrVqxduxbnz59HREQETpw4geDgYABAQEAALC0t4e/vj4MHDyIzMxPJyckICQnBlStXtBJzTbfh5+cHU1NTLFiwAKNHj9ZCxERERFRXMKGtxtq1a3H//n14enoiNDQUCxYseCbtBgYG4t69e+jcuTOmTJmCqVOnYvz48dUeExUVhcTERLi7u2PDhg1ISEhA69atAQCGhob4+eef4eDggMGDB8PNzQ1BQUEoLCzUWm9qTbfRoEEDjBo1CqWlpSo90URERESSeHiQKNU6X19fdOjQQaNX20qShB07dqi8urYuCgoKwt9//41vvvlG42Pz8/MfzHYQugUNFIY1EB0REVH9lLVwQI3VXf77Oy8vr9oOMo6hpedeXl4e/ve//2HTpk1PlMwSERFR3caE9hnKzs6WhwFU5ty5c88wmmfL2Ni4yn337t2DUqmsdN/u3bsRHh6OEydOYOLEiXj55ZdrKkQiIiLSUUxonyE7O7tqZ0qws7N7opcQ6MKokerOu7qEtmnTpk/9YgYiIiKq2ziGluoFdcfgEBER0fND3d/fnOWAiIiIiHQaE1oiIiIi0mlMaImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaImIiIhIpzWs7QCInqW2EXvQQGFY22EQEREha+GA2g6hzmAPLRERERHpNCa0RERERKTTmNASERERkU5jQktEREREOo0JLRERERHpNCa0WuDo6IjY2Fit1jlq1CgMGjRIXvf19UVoaKhW2yAiIiKqCzhtlxacPHkSRkZGNdrG9u3boa+vX6Nt6AJJkrBjxw6VZJ+IiIjqNya0WmBlZVXjbVhYWNR4G7WltLQUkiShQQP+wYCIiIg091xlEL6+vggJCcHs2bNhYWEBGxsbREZGAgCysrIgSRJSUlLk8rdv34YkSUhOTgYAJCcnQ5Ik7NmzBx4eHlAqlejduzdyc3Oxe/duuLm5wdTUFG+//Tbu3r2rdkzBwcEIDg6GmZkZLC0tER4eDiGEXObRIQeSJCEuLg79+/eHUqmEk5MTtm3bplLv5cuXMXToUJibm8PCwgL+/v7IysqqNo6Hhxw4Ojrio48+wpgxY2BiYgIHBwesXr36qdrQxNq1a9GmTRsoFArY2toiODhY3rd48WK0a9cORkZGsLe3x+TJk1FQUCDvj4+Ph7m5Ob755hu0bt0aCoUC2dnZOHnyJF5++WVYWlrCzMwMPXv2xKlTp1TOGQDeeOMNSJIkrxMREVH99lwltACwfv16GBkZ4fjx41i0aBHmz5+PvXv3alRHZGQkVqxYgSNHjshJXWxsLDZt2oSkpCT88MMPWL58uUYxNWzYECdOnMDSpUuxePFirFmzptpjwsPDMWTIEKSmpiIgIADDhw9HWloaAKCkpAR+fn4wMTHBwYMHcfjwYRgbG6Nfv34oLi5WO66YmBh4eXnh9OnTmDx5MiZNmoT09HSttlGZuLg4TJkyBePHj8f//vc/fPPNN3B2dpb3N2jQAMuWLcPZs2exfv16/PTTT5g9e7ZKHXfv3sXHH3+MNWvW4OzZs7C2tsadO3cwcuRIHDp0CMeOHYOLiwteffVV3LlzB8CDoR0AsG7dOuTk5MjrlSkqKkJ+fr7KQkRERHXTczfkwN3dHREREQAAFxcXrFixAvv27YOLi4vadSxYsADdunUDAAQFBWHOnDnIyMiAk5MTAODNN9/E/v37ERYWplZ99vb2WLJkCSRJgqurK/73v/9hyZIlGDduXJXHvPXWWxg7diwA4IMPPsDevXuxfPlyrFy5Eps3b0ZZWRnWrFkDSZIAPEjSzM3NkZycjFdeeUWtuF599VVMnjwZABAWFoYlS5Zg//79cHV11VoblVmwYAFmzJiBqVOnyts6deok//xoT/KCBQswceJErFy5Ut5eUlKClStXon379vK23r17q7SzevVqmJub48CBA3jttdfkoR3m5uawsbGpNsbo6GhERUU90fkRERGRbnnuemjd3d1V1m1tbZGbm/vEdTRp0gSGhoZyMlu+TZM6vb295aQQAHx8fHDhwgWUlpZWeYyPj0+F9fIe2tTUVFy8eBEmJiYwNjaGsbExLCwsUFhYiIyMDLXjevg8JUmCjY2NfF7aauNRubm5+Ouvv9CnT58qy/z444/o06cPmjZtChMTE7z77ru4ceOGyjAPAwODCvf62rVrGDduHFxcXGBmZgZTU1MUFBQgOztb4zjnzJmDvLw8ebl8+bLGdRAREZFueO56aB99kl+SJJSVlckPDD08drWkpOSxdUiSVGWdtaWgoACenp7YuHFjhX2aPGBW3Xlpq41HKZXKavdnZWXhtddew6RJk/Dhhx/CwsIChw4dQlBQEIqLi2FoaCjX8/B/EgBg5MiRuHHjBpYuXYrmzZtDoVDAx8fniYZIKBQKKBQKjY8jIiIi3fPcJbRVKU/CcnJy4OHhAQAqD4jVpOPHj6usl4/v1NPTq/KYY8eOITAwUGW9PO6OHTti8+bNsLa2hqmpaY3EXFNtmJiYwNHREfv27UOvXr0q7P/1119RVlaGmJgY+T8hW7ZsUavuw4cPY+XKlXj11VcBPHio7fr16ypl9PX1q+0ZJyIiovrnuRtyUBWlUglvb28sXLgQaWlpOHDgAObNm/dM2s7Ozsb06dORnp6OhIQELF++XGX8aGW2bt2KtWvX4vz584iIiMCJEyfkmQACAgJgaWkJf39/HDx4EJmZmUhOTkZISAiuXLmilZhrso3IyEjExMRg2bJluHDhAk6dOiU/ZOfs7IySkhIsX74cf/zxB7788kt89tlnatXr4uKCL7/8EmlpaTh+/DgCAgIq9AiXJ9NXr17FrVu3nuo8iIiIqG7QmYQWeDBV1P379+Hp6YnQ0FAsWLDgmbQbGBiIe/fuoXPnzpgyZQqmTp2K8ePHV3tMVFQUEhMT4e7ujg0bNiAhIQGtW7cGABgaGuLnn3+Gg4MDBg8eDDc3NwQFBaGwsFBrvak12cbIkSMRGxuLlStXok2bNnjttddw4cIFAED79u2xePFifPzxx2jbti02btyI6Ohoter94osvcOvWLXTs2BHvvvsuQkJCYG1trVImJiYGe/fuhb29vdzjTURERPWbJB4elEoV+Pr6okOHDhq92pZvs3r+5Ofnw8zMDPahW9BAYVjb4RARESFr4YDaDuG5V/77Oy8vr9oOOZ3qoSUiIiIiepTOPBRWE7Kzs+VhAJU5d+7cM4zm2TI2Nq5y371796qczWD37t3o3r17TYVFREREpLF6ndDa2dlVO1OCnZ2d/FpdTejCKI7qzru6hLZp06Y1FBERERHRk+EYWqoX1B2DQ0RERM8PjqElIiIionqBCS0RERER6TQmtERERESk05jQEhEREZFOY0JLRERERDqNCS0RERER6TQmtERERESk05jQEhEREZFOY0JLRERERDqNCS0RERER6TQmtERERESk05jQEhEREZFOY0JLRERERDqNCS0RERER6TQmtERERESk05jQEhEREZFOY0JLRERERDqNCS0RERER6TQmtERERESk05jQEhEREZFOa1jbARA9S20j9qCBwrC2wyAiIh2TtXBAbYdA1WAPLRERERHpNCa0RERERKTTmNASERERkU5jQktEREREOo0JrQ5ydHREbGysVuscNWoUBg0aJK/7+voiNDRUq20QERER1QTOcqCDTp48CSMjoxptY/v27dDX16/RNoiIiIi0gQmtDrKysqrxNiwsLGq8DSIiIiJtqNdDDnx9fRESEoLZs2fDwsICNjY2iIyMBABkZWVBkiSkpKTI5W/fvg1JkpCcnAwASE5OhiRJ2LNnDzw8PKBUKtG7d2/k5uZi9+7dcHNzg6mpKd5++23cvXtX7ZiCg4MRHBwMMzMzWFpaIjw8HEIIucyjQw4kSUJcXBz69+8PpVIJJycnbNu2TaXey5cvY+jQoTA3N4eFhQX8/f2RlZVVbRwPDzlwdHTERx99hDFjxsDExAQODg5YvXr1U7WhjsLCQrRp0wbjx4+Xt2VkZMDExARr1659qrqJiIiobqjXCS0ArF+/HkZGRjh+/DgWLVqE+fPnY+/evRrVERkZiRUrVuDIkSNyUhcbG4tNmzYhKSkJP/zwA5YvX65RTA0bNsSJEyewdOlSLF68GGvWrKn2mPDwcAwZMgSpqakICAjA8OHDkZaWBgAoKSmBn58fTExMcPDgQRw+fBjGxsbo168fiouL1Y4rJiYGXl5eOH36NCZPnoxJkyYhPT1dq208qlGjRti4cSPWr1+Pr7/+GqWlpXjnnXfw8ssvY8yYMVUeV1RUhPz8fJWFiIiI6qZ6n9C6u7sjIiICLi4uCAwMhJeXF/bt26dRHQsWLEC3bt3g4eGBoKAgHDhwAHFxcfDw8ED37t3x5ptvYv/+/WrXZ29vjyVLlsDV1RUBAQF47733sGTJkmqPeeuttzB27Fi0bNkSH3zwAby8vOQkevPmzSgrK8OaNWvQrl07uLm5Yd26dcjOzpZ7m9Xx6quvYvLkyXB2dkZYWBgsLS3l89JWG5Xp0KEDFixYgLFjxyI0NBSXLl3C559/Xu0x0dHRMDMzkxd7e/unioGIiIieX0xo3d1V1m1tbZGbm/vEdTRp0gSGhoZwcnJS2aZJnd7e3pAkSV738fHBhQsXUFpaWuUxPj4+FdbLe2hTU1Nx8eJFmJiYwNjYGMbGxrCwsEBhYSEyMjLUjuvh85QkCTY2NvJ5aauNqsyYMQMtW7bEihUrsHbtWjRu3Lja8nPmzEFeXp68XL58+aljICIioudTvX8o7NEn+SVJQllZGRo0eJDrPzx2taSk5LF1SJJUZZ21paCgAJ6enti4cWOFfZo8YFbdeWmrjark5ubi/Pnz0NPTw4ULF9CvX79qyysUCigUiqdul4iIiJ5/9b6HtirlSVhOTo687eEHxGrS8ePHVdaPHTsGFxcX6OnpVXnMsWPHKqy7ubkBADp27IgLFy7A2toazs7OKouZmZlWYq7pNsaMGYN27dph/fr1CAsLk3ufiYiIiJjQVkGpVMLb2xsLFy5EWloaDhw4gHnz5j2TtrOzszF9+nSkp6cjISEBy5cvx9SpU6s9ZuvWrVi7di3Onz+PiIgInDhxAsHBwQCAgIAAWFpawt/fHwcPHkRmZiaSk5MREhKCK1euaCXmmmzj008/xdGjR7F+/XoEBARg0KBBCAgIeKqHzYiIiKjuYEJbjbVr1+L+/fvw9PREaGgoFixY8EzaDQwMxL1799C5c2dMmTIFU6dOVZm2qjJRUVFITEyEu7s7NmzYgISEBLRu3RoAYGhoiJ9//hkODg4YPHgw3NzcEBQUhMLCQpiammol5ppq4/fff8esWbOwcuVK+cGulStX4vr16wgPD9dK7ERERKTbJPHwIFGqdb6+vujQoYNGr7aVJAk7duxQeXUtqcrPz38w20HoFjRQGNZ2OEREpGOyFg6o7RDqpfLf33l5edV2kLGHloiIiIh0Wr2f5eBZys7OlocBVObcuXPPMJpny9jYuMp99+7dg1KprHTf7t270b1795oKi4iIiOoAJrTPkJ2dXbUzJdjZ2T3RSwh0YdRIdeddXULbtGnTGoqIiIiI6gqOoaV6Qd0xOERERPT84BhaIiIiIqoXmNASERERkU5jQktEREREOo0JLRERERHpNCa0RERERKTTmNASERERkU5jQktEREREOo0JLRERERHpNCa0RERERKTTmNASERERkU5jQktEREREOo0JLRERERHpNCa0RERERKTTmNASERERkU5jQktEREREOo0JLRERERHpNCa0RERERKTTmNASERERkU5jQktEREREOo0JLRERERHptIa1HQDRs9Q2Yg8aKAxrOwwiojota+GA2g6B6hn20BIRERGRTmNCS0REREQ6jQktEREREek0JrREREREpNOeKKG9f/8+fvzxR6xatQp37twBAPz1118oKCjQanBUNUdHR8TGxmq1zlGjRmHQoEHyuq+vL0JDQ7XaBhEREZG2aTzLwaVLl9CvXz9kZ2ejqKgIL7/8MkxMTPDxxx+jqKgIn332WU3ESY84efIkjIyMarSN7du3Q19fv0bbICIiInpaGvfQTp06FV5eXrh16xaUSqW8/Y033sC+ffu0GhxVzcrKCoaGNTv9lIWFBUxMTGq0DSIiIqKnpXFCe/DgQcybNw8GBgYq2x0dHfHnn39qLbBnxdfXFyEhIZg9ezYsLCxgY2ODyMhIAEBWVhYkSUJKSopc/vbt25AkCcnJyQCA5ORkSJKEPXv2wMPDA0qlEr1790Zubi52794NNzc3mJqa4u2338bdu3fVjik4OBjBwcEwMzODpaUlwsPDIYSQyzw65ECSJMTFxaF///5QKpVwcnLCtm3bVOq9fPkyhg4dCnNzc1hYWMDf3x9ZWVnVxvHwkANHR0d89NFHGDNmDExMTODg4IDVq1c/VRvqkiSpwuLo6PjU9RIREZHu0zihLSsrQ2lpaYXtV65c0dnevPXr18PIyAjHjx/HokWLMH/+fOzdu1ejOiIjI7FixQocOXJETupiY2OxadMmJCUl4YcffsDy5cs1iqlhw4Y4ceIEli5disWLF2PNmjXVHhMeHo4hQ4YgNTUVAQEBGD58ONLS0gAAJSUl8PPzg4mJCQ4ePIjDhw/D2NgY/fr1Q3FxsdpxxcTEwMvLC6dPn8bkyZMxadIkpKena7WNyuTk5MjLxYsX4ezsjB49elRZvqioCPn5+SoLERER1U0aJ7SvvPJKhZ7BgoICRERE4NVXX9VmbM+Mu7s7IiIi4OLigsDAQHh5eWk8fGLBggXo1q0bPDw8EBQUhAMHDiAuLg4eHh7o3r073nzzTezfv1/t+uzt7bFkyRK4uroiICAA7733HpYsWVLtMW+99RbGjh2Lli1b4oMPPoCXl5ecRG/evBllZWVYs2YN2rVrBzc3N6xbtw7Z2dlyb7M6Xn31VUyePBnOzs4ICwuDpaWlfF7aaqMyNjY2sLGxQZMmTTBr1iyYmZlh1apVVZaPjo6GmZmZvNjb2z9V+0RERPT80jihjYmJweHDh9G6dWsUFhbi7bfflocbfPzxxzURY41zd3dXWbe1tUVubu4T19GkSRMYGhrCyclJZZsmdXp7e0OSJHndx8cHFy5cqLR3/OEyj66X99Cmpqbi4sWLMDExgbGxMYyNjWFhYYHCwkJkZGSoHdfD5ylJEmxsbOTz0lYb1fn3v/+No0eP4uuvv1YZw/2oOXPmIC8vT14uX76slfaJiIjo+aPxLAfNmjVDamoqEhMT8dtvv6GgoABBQUEICAioNsF4nj36JL8kSSgrK0ODBg/y/YfHrpaUlDy2DkmSqqyzthQUFMDT0xMbN26ssM/Kykrteqo7L221UZWvvvoKS5YsQXJyMpo2bVptWYVCAYVC8dRtEhER0fNP44QWABo2bIh33nlH27E8d8qTsJycHHh4eACAygNiNen48eMq68eOHYOLiwv09PSqPObYsWMIDAxUWS+Pu2PHjti8eTOsra1hampaIzHXZBtHjx7F2LFjsWrVKnh7e2u1biIiItJtT5TQ/vXXXzh06BByc3Mr9DqGhIRoJbDngVKphLe3NxYuXIgWLVogNzcX8+bNeyZtZ2dnY/r06ZgwYQJOnTqF5cuXIyYmptpjtm7dCi8vL7z00kvYuHEjTpw4gS+++AIAEBAQgE8++QT+/v6YP38+mjVrhkuXLmH79u2YPXs2mjVr9tQx11QbV69exRtvvIHhw4fDz88PV69eBQDo6elppeeXiIiIdJvGCW18fDwmTJgAAwMDNG7cWGWcpyRJdSqhBYC1a9ciKCgInp6ecHV1xaJFi/DKK6/UeLuBgYG4d+8eOnfuDD09PUydOhXjx4+v9pioqCgkJiZi8uTJsLW1RUJCAlq3bg0AMDQ0xM8//4ywsDAMHjwYd+7cQdOmTdGnTx+t9abWVBu///47rl27hvXr12P9+vXy9ubNm2tlSjAiIiLSbZJ4eICoGuzt7TFx4kTMmTNHHmNK2uXr64sOHTpo9GpbSZKwY8cOlVfX0v/Jz89/MNtB6BY0UNTsCymIiOq7rIUDajsEqiPKf3/n5eVV2zmmcUZ69+5dDB8+nMksERERET0XNM5Kg4KCsHXr1pqIpV7Izs6Wp7SqbMnOzq7tEGtMdeetp6dX5b6DBw/WduhERET0HNN4yEFpaSlee+013Lt3D+3ataswjdPixYu1GmBdc//+/WrHfTo6OqJhwyd6Vu+5d/HixSr33bt3r8pp35o2bfrUU8JxyAER0bPDIQekLeoOOdA4c4qOjsaePXvg6uoKABUeCqPqNWzYEM7OzrUdRq14Hs77TJRfjU1bRkRERLVD44Q2JiYGa9euxahRo2ogHCIiIiIizWg8hlahUKBbt241EQsRERERkcY0TminTp2K5cuX10QsREREREQa03jIwYkTJ/DTTz/h22+/RZs2bSo8FLZ9+3atBUdERERE9DgaJ7Tm5uYYPHhwTcRCRERERKQxjRPadevW1UQcRERERERPhK/7IiIiIiKd9kQz+G/btg1btmxBdnY2iouLVfadOnVKK4EREREREalD4x7aZcuWYfTo0WjSpAlOnz6Nzp07o3Hjxvjjjz/Qv3//moiRiIiIiKhKGie0K1euxOrVq7F8+XIYGBhg9uzZ2Lt3L0JCQpCXl1cTMRIRERERVUnjhDY7Oxtdu3YFACiVSty5cwcA8O677yIhIUG70RERERERPYbGCa2NjQ1u3rwJAHBwcMCxY8cAAJmZmRBCaDc6IiIiIqLH0Dih7d27N7755hsAwOjRozFt2jS8/PLLGDZsGN544w2tB0hEREREVB1JaNitWlZWhrKyMjRs+GCChMTERBw5cgQuLi6YMGECDAwMaiRQoqeRn58PMzMz5OXlwdTUtLbDISIiIjWo+/tb44SWSBcxoSUiItI96v7+fqJ5aG/fvo0TJ04gNzcXZWVlKvsCAwOfpEoiIiIioieicUK7a9cuBAQEoKCgAKamppAkSd4nSRITWiIiIiJ6pjR+KGzGjBkYM2YMCgoKcPv2bdy6dUteymc/ICIiIiJ6VjROaP/880+EhITA0NCwJuIhIiIiItKIxgmtn58ffvnll5qIhYiIiIhIYxqPoR0wYABmzZqFc+fOoV27dtDX11fZ//rrr2stOCJtaxuxBw0U/OsCEdUdWQsH1HYIRLVO44R23LhxAID58+dX2CdJEkpLS58+KiIiIiIiNWmc0D46TRcRERERUW3SeAwtEREREdHzhAktEREREek0JrRqysrKgiRJSElJ0Wq9jo6OiI2NldclScLOnTu12kZNSE5OhiRJuH379lPXpSvnTERERM+nOpnQRkZGokOHDlqt097eHjk5OWjbtq1W631UTk4O+vfvX6NtPG8ePuea+o8DERER1V0aPxRWX+np6cHGxqbG23kWbTwviouLYWBgUK/OmYiIiLTviXpoMzIyMG/ePIwYMQK5ubkAgN27d+Ps2bNaC6ysrAyLFi2Cs7MzFAoFHBwc8OGHHwIAwsLC0LJlSxgaGsLJyQnh4eEoKSkBAMTHxyMqKgqpqamQJAmSJCE+Pv6x7UmShLi4OPTv3x9KpRJOTk7Ytm2bvP/RnsPyP7knJSXB3d0djRo1gre3N86cOaNS76FDh9C9e3colUrY29sjJCQE//zzT7VxlP/5vbzN7du3o1evXjA0NET79u1x9OjRp2pDXUVFRQgLC4O9vT0UCgWcnZ3xxRdfVFr2xo0bGDFiBJo2bQpDQ0O0a9cOCQkJKmV8fX0RHByM0NBQWFpaws/Pr8I5t2jRAgDg4eEBSZLg6+uLn3/+Gfr6+rh69apKfaGhoejevftTnycRERHpNo0T2gMHDqBdu3Y4fvw4tm/fjoKCAgBAamoqIiIitBbYnDlzsHDhQoSHh+PcuXPYtGkTmjRpAgAwMTFBfHw8zp07h6VLl+Lzzz/HkiVLAADDhg3DjBkz0KZNG+Tk5CAnJwfDhg1Tq83w8HAMGTIEqampCAgIwPDhw5GWllbtMbNmzUJMTAxOnjwJKysrDBw4UE6uMzIy0K9fPwwZMgS//fYbNm/ejEOHDiE4OFijazF37lzMnDkTKSkpaNmyJUaMGIH79+9rtY3KBAYGIiEhAcuWLUNaWhpWrVoFY2PjSssWFhbC09MTSUlJOHPmDMaPH493330XJ06cUCm3fv16GBgY4PDhw/jss88q1FNe/scff0ROTg62b9+OHj16wMnJCV9++aVcrqSkBBs3bsSYMWMqjaeoqAj5+fkqCxEREdVNkhBCaHKAj48P3nrrLUyfPh0mJiZITU2Fk5MTTpw4gcGDB+PKlStPHdSdO3dgZWWFFStWYOzYsY8t/5///AeJiYnyK3kjIyOxc+dOjcZhSpKEiRMnIi4uTt7m7e2Njh07YuXKlcjKykKLFi1w+vRpdOjQAcnJyejVqxcSExPlhPnmzZto1qwZ4uPjMXToUIwdOxZ6enpYtWqVXOehQ4fQs2dP/PPPP2jUqBEcHR0RGhqK0NBQOY4dO3Zg0KBBcptr1qxBUFAQAODcuXNo06YN0tLS0KpVK7XaeBLnz5+Hq6sr9u7di759+1bYX37+t27dgrm5eaV1vPbaa2jVqhX+85//AHjQQ5ufn49Tp06plKvsnMuvc7lFixbJ/4kBgO3bt2PkyJG4evUqjIyMKrQdGRmJqKioCtvtQ7fwTWFEVKfwTWFUl+Xn58PMzAx5eXkwNTWtspzGPbT/+9//8MYbb1TYbm1tjevXr2taXaXS0tJQVFSEPn36VLp/8+bN6NatG2xsbGBsbIx58+YhOzv7qdv18fGpsP64HtqHj7GwsICrq6t8TGpqKuLj42FsbCwvfn5+KCsrQ2Zmptpxubu7yz/b2toCgDzUQ1ttPColJQV6enro2bOnWuVLS0vxwQcfoF27drCwsICxsTH27NlT4b54eno+UTyjRo3CxYsXcezYMQCQ/9NQWTILPOjhz8vLk5fLly8/UbtERET0/NP4oTBzc3Pk5OTIYx3LnT59Gk2bNtVKUEqlssp9R48eRUBAAKKiouDn5wczMzMkJiYiJiZGK21rU0FBASZMmICQkJAK+xwcHNSuR19fX/5ZkiQA//fGNm218ajq7kFlPvnkEyxduhSxsbFo164djIyMEBoaiuLiYpVyVSWgj2NtbY2BAwdi3bp1aNGiBXbv3o3k5OQqyysUCigUiidqi4iIiHSLxgnt8OHDERYWhq1bt0KSJJSVleHw4cOYOXMmAgMDtRKUi4sLlEol9u3bV2HIwZEjR9C8eXPMnTtX3nbp0iWVMgYGBigtLdW43WPHjqmcw7Fjx+Dh4fHYY8oTx1u3buH8+fNwc3MDAHTs2BHnzp2Ds7OzxrGoq6baaNeuHcrKynDgwIFKhxw86vDhw/D398c777wD4EHCff78ebRu3Vqjdg0MDACg0vs3duxYjBgxAs2aNcOLL76Ibt26aVQ3ERER1U0aDzn46KOP0KpVK9jb26OgoACtW7dGjx490LVrV8ybN08rQTVq1AhhYWGYPXs2NmzYgIyMDBw7dgxffPEFXFxckJ2djcTERGRkZGDZsmXYsWOHyvGOjo7IzMxESkoKrl+/jqKiIrXa3bp1K9auXYvz588jIiICJ06ceOzDVfPnz8e+fftw5swZjBo1CpaWlhg0aBCAB7MxHDlyBMHBwUhJScGFCxfw9ddfa+WBrXI11YajoyNGjhyJMWPGYOfOncjMzERycjK2bNlSaXkXFxfs3bsXR44cQVpaGiZMmIBr165p3K61tTWUSiW+//57XLt2DXl5efI+Pz8/mJqaYsGCBRg9evQTnxsRERHVLRoltEIIXL16FcuWLcMff/yBb7/9Fl999RV+//13fPnll9DT09NaYOHh4ZgxYwbef/99uLm5YdiwYcjNzcXrr7+OadOmITg4GB06dMCRI0cQHh6ucuyQIUPQr18/9OrVC1ZWVhWmj6pKVFQUEhMT4e7ujg0bNiAhIeGxPYwLFy7E1KlT4enpiatXr2LXrl1yL6O7uzsOHDiA8+fPo3v37vDw8MD7778POzu7J7solajJNuLi4vDmm29i8uTJaNWqFcaNG1fldGDz5s1Dx44d4efnB19fX9jY2MiJvSYaNmyIZcuWYdWqVbCzs4O/v7+8r0GDBhg1ahRKS0u19tcAIiIi0n0azXJQVlaGRo0a4ezZs3BxcanJuJ65h5+0V4c6T/mT9gUFBeHvv//GN998o9Fx5U9JcpYDIqprOMsB1WXqznKg0RjaBg0awMXFBTdu3KhzCS093/Ly8vC///0PmzZt0jiZJSIiorpN4zG0CxcuxKxZsyq8Eet5tnHjRpVprR5e2rRpU9vh1ZiDBw9Wed56enpV7qvq5Qm1yd/fH6+88gomTpyIl19+ubbDISIioueIxi9WeOGFF3D37l3cv38fBgYGFaZ3unnzplYD1IY7d+5U+YCSvr4+mjdv/owjejbu3buHP//8s8p91U3NVZMzM9QGDjkgorqKQw6oLquRIQcAEBsb+zRx1QoTExOYmJjUdhjPnFKprHOJ6dM6E+VX7ReCiIiIdI/GCe3IkSNrIg4iIiIioieicUL7uFfMPs3bqYiIiIiINKVxQuvo6Ci/frUyT/KGLiIiIiKiJ6VxQnv69GmV9ZKSEpw+fRqLFy/Ghx9+qLXAiIiIiIjUoXFC2759+wrbvLy8YGdnh08++QSDBw/WSmBEREREROrQeB7aqri6uuLkyZPaqo6IiIiISC0a99Dm5+errAshkJOTg8jISL49jIiIiIieOY0TWnNz8woPhQkhYG9vj8TERK0FRkRERESkDo0T2v3796usN2jQAFZWVnB2dkbDhhpXR0RERET0VDTOQCVJQteuXSskr/fv38fPP/+MHj16aC04IiIiIqLH0fihsF69euHmzZsVtufl5aFXr15aCYqIiIiISF0aJ7RCiEpfrHDjxg0YGRlpJSgiIiIiInWpPeSgfH5ZSZIwatQoKBQKeV9paSl+++03dO3aVfsREhERERFVQ+2E1szMDMCDHloTExMolUp5n4GBAby9vTFu3DjtR0hEREREVA21E9p169YBABwdHTFz5kwOLyAiIiKi54IkhBC1HQRRTcvPz4eZmRny8vJgampa2+EQERGRGtT9/f1EE8du27YNW7ZsQXZ2NoqLi1X2nTp16kmqJCIiIiJ6IhrPcrBs2TKMHj0aTZo0wenTp9G5c2c0btwYf/zxB/r3718TMRIRERERVUnjhHblypVYvXo1li9fDgMDA8yePRt79+5FSEgI8vLyaiJGIiIiIqIqaZzQZmdny9NzKZVK3LlzBwDw7rvvIiEhQbvRERERERE9hsZjaG1sbHDz5k00b94cDg4OOHbsGNq3b4/MzEzw+TJ63rWN2IMGCsPaDoNIZ2UtHFDbIRARVaBxD23v3r3xzTffAABGjx6NadOm4eWXX8awYcPwxhtvaD1AIiIiIqLqaNxDu3r1apSVlQEApkyZgsaNG+PIkSN4/fXXMWHCBK0HSERERERUHY0T2gYNGqBBg//r2B0+fDiGDx+u1aCIiIiIiNSl8ZADADh48CDeeecd+Pj44M8//wQAfPnllzh06JBWgyMiIiIiehyNE9r//ve/8PPzg1KpxOnTp1FUVAQAyMvLw0cffaT1AHVJVlYWJElCSkqKVut1dHREbGysvC5JEnbu3KnVNmqLr68vQkNDazsMIiIi0mEaJ7QLFizAZ599hs8//xz6+vry9m7duunUW8IiIyPRoUMHrdZpb2+PnJwctG3bVqv1PionJ4cvsSAiIiL6/zQeQ5ueno4ePXpU2G5mZobbt29rIyadpaenBxsbmxpv51m0oStKS0shSZLKuG4iIiKqXzTOAmxsbHDx4sUK2w8dOgQnJyetBKWusrIyLFq0CM7OzlAoFHBwcMCHH34IAAgLC0PLli1haGgIJycnhIeHo6SkBAAQHx+PqKgopKamQpIkSJKE+Pj4x7YnSRLi4uLQv39/KJVKODk5Ydu2bfL+R4ccJCcnQ5IkJCUlwd3dHY0aNYK3tzfOnDmjUu+hQ4fQvXt3KJVK2NvbIyQkBP/880+1cZQPOShvc/v27ejVqxcMDQ3Rvn17HD169KnaUFdRURHCwsJgb28PhUIBZ2dnfPHFF/L+AwcOoHPnzlAoFLC1tcW//vUv3L9/v8r6bt26hcDAQLzwwgswNDRE//79ceHCBXl/fHw8zM3N8c0336B169ZQKBTIzs5+6vMgIiIi3aVxQjtu3DhMnToVx48fhyRJ+Ouvv7Bx40bMnDkTkyZNqokYqzRnzhwsXLgQ4eHhOHfuHDZt2oQmTZoAAExMTBAfH49z585h6dKl+Pzzz7FkyRIAwLBhwzBjxgy0adMGOTk5yMnJwbBhw9RqMzw8HEOGDEFqaioCAgIwfPhwpKWlVXvMrFmzEBMTg5MnT8LKygoDBw6Uk+uMjAz069cPQ4YMwW+//YbNmzfj0KFDCA4O1uhazJ07FzNnzkRKSgpatmyJESNGyImjttqoTGBgIBISErBs2TKkpaVh1apVMDY2BgD8+eefePXVV9GpUyekpqYiLi4OX3zxBRYsWFBlfaNGjcIvv/yCb775BkePHoUQAq+++qp8vQDg7t27+Pjjj7FmzRqcPXsW1tbWFeopKipCfn6+ykJERER1kyQ0fL2XEAIfffQRoqOjcffuXQCAQqHAzJkz8cEHH9RIkJW5c+cOrKyssGLFCowdO/ax5f/zn/8gMTERv/zyC4AHY2h37typ0QNckiRh4sSJiIuLk7d5e3ujY8eOWLlyJbKystCiRQucPn0aHTp0QHJyMnr16oXExEQ5Yb558yaaNWuG+Ph4DB06FGPHjoWenh5WrVol13no0CH07NkT//zzDxo1agRHR0eEhobKD09JkoQdO3Zg0KBBcptr1qxBUFAQAODcuXNo06YN0tLS0KpVK7XaeBLnz5+Hq6sr9u7di759+1bYP3fuXPz3v/9FWloaJEkCAKxcuRJhYWHIy8tDgwYN4Ovriw4dOiA2NhYXLlxAy5YtcfjwYfn1yjdu3IC9vT3Wr1+Pt956C/Hx8Rg9ejRSUlLQvn37KmOLjIxEVFRUhe32oVv4pjCip8A3hRHRs5Sfnw8zMzPk5eXB1NS0ynJq9dD+9ttv8ssUJEnC3LlzcfPmTZw5cwbHjh3D33///UyTWQBIS0tDUVER+vTpU+n+zZs3o1u3brCxsYGxsTHmzZunlT9N+/j4VFh/XA/tw8dYWFjA1dVVPiY1NRXx8fEwNjaWFz8/P5SVlSEzM1PtuNzd3eWfbW1tAQC5ublabeNRKSkp0NPTQ8+ePSvdn5aWBh8fHzmZBR48PFhQUIArV65UWr5hw4bo0qWLvK1x48Yq1wsADAwMVM63MnPmzEFeXp68XL58WdPTIyIiIh2h1kNhHh4eyMnJgbW1NZycnHDy5Ek0btwYrVu3run4qqRUKqvcd/ToUQQEBCAqKgp+fn4wMzNDYmIiYmJinmGE6ikoKMCECRMQEhJSYZ+Dg4Pa9Tw840R5Aln+nxBttfGo6u5BTVIqlSpJcmUUCgUUCsUzioiIiIhqk1o9tObm5nJPXlZWlpwo1SYXFxcolUrs27evwr4jR46gefPmmDt3Lry8vODi4oJLly6plDEwMEBpaanG7R47dqzCupubm9rH3Lp1C+fPn5eP6dixI86dOwdnZ+cKi4GBgcbxVaam2mjXrh3Kyspw4MCBSve7ubnJ42DLHT58GCYmJmjWrFml5e/fv4/jx4/L227cuIH09PRa/c8TERERPd/U6qEdMmQIevbsCVtbW0iSBC8vL+jp6VVa9o8//tBqgFVp1KgRwsLCMHv2bBgYGKBbt274+++/cfbsWbi4uCA7OxuJiYno1KkTkpKSsGPHDpXjHR0dkZmZiZSUFDRr1gwmJiZq9eht3boVXl5eeOmll7Bx40acOHFC5an+ysyfPx+NGzdGkyZNMHfuXFhaWmLQoEEAHszG4O3tjeDgYIwdOxZGRkY4d+4c9u7dixUrVjzx9XlYTbXh6OiIkSNHYsyYMVi2bBnat2+PS5cuITc3F0OHDsXkyZMRGxuL9957D8HBwUhPT0dERASmT59e6TRbLi4u8Pf3x7hx47Bq1SqYmJjgX//6F5o2bQp/f/+nuQRERERUh6mV0K5evRqDBw/GxYsXERISgnHjxsHExKSmY3us8PBwNGzYEO+//z7++usv2NraYuLEiQgKCsK0adMQHByMoqIiDBgwAOHh4YiMjJSPHTJkiDzV1e3bt7Fu3TqMGjXqsW1GRUUhMTERkydPhq2tLRISEh7be7hw4UJMnToVFy5cQIcOHbBr1y65Z9Td3R0HDhzA3Llz0b17dwgh8OKLL6o964I6arKNuLg4/Pvf/8bkyZNx48YNODg44N///jcAoGnTpvjuu+8wa9YstG/fHhYWFggKCsK8efOqrG/dunWYOnUqXnvtNRQXF6NHjx747rvvVIZUEBERET1M41kORo8ejWXLlj0XCe2z9vDsAuoon+Xg1q1bMDc3r9HYqHrlT0lylgOip8NZDojoWVJ3lgON3xS2bt26pwqMiIiIiEib+L7Q/2/jxo0q01o9vLRp06a2w6sxBw8erPK89fT0qtxX/vIEIiIiotqmcQ9tXfX666+rzH/6sPLxmxqOzoCvr6/GxzxrXl5eVb5c4t69e7U2NRcRERGRujQeQ0uki9Qdg0NERETPD62+KYyIiIiI6HnFhJaIiIiIdBoTWiIiIiLSaUxoiYiIiEinMaElIiIiIp3GhJaIiIiIdBoTWiIiIiLSaUxoiYiIiEinMaElIiIiIp3GhJaIiIiIdBoTWiIiIiLSaUxoiYiIiEinMaElIiIiIp3GhJaIiIiIdBoTWiIiIiLSaUxoiYiIiEinMaElIiIiIp3GhJaIiIiIdBoTWiIiIiLSaUxoiYiIiEinNaztAIiepbYRe9BAYSivZy0cUIvREBERkTawh5aIiIiIdBoTWiIiIiLSaUxoiYiIiEinMaElIiIiIp3GhLYWZWVlQZIkpKSkaLVeR0dHxMbGyuuSJGHnzp1abYOIiIjoecGEVk2RkZHo0KGDVuu0t7dHTk4O2rZtq9V6H5WTk4P+/fvXaBvPSk3cByIiItJtnLarFunp6cHGxqbG23kWbRARERHVlnrVQ1tWVoZFixbB2dkZCoUCDg4O+PDDDwEAYWFhaNmyJQwNDeHk5ITw8HCUlJQAAOLj4xEVFYXU1FRIkgRJkhAfH//Y9iRJQlxcHPr37w+lUgknJyds27ZN3v/okIPk5GRIkoSkpCS4u7ujUaNG8Pb2xpkzZ1TqPXToELp37w6lUgl7e3uEhITgn3/+qTaO8iEH5W1u374dvXr1gqGhIdq3b4+jR48+VRvqKioqQlhYGOzt7aFQKODs7IwvvvgCwIPrbG5urlJ+586dkCRJ3v8k94GIiIjqtnqV0M6ZMwcLFy5EeHg4zp07h02bNqFJkyYAABMTE8THx+PcuXNYunQpPv/8cyxZsgQAMGzYMMyYMQNt2rRBTk4OcnJyMGzYMLXaDA8Px5AhQ5CamoqAgAAMHz4caWlp1R4za9YsxMTE4OTJk7CyssLAgQPl5DojIwP9+vXDkCFD8Ntvv2Hz5s04dOgQgoODNboWc+fOxcyZM5GSkoKWLVtixIgRuH//vlbbqExgYCASEhKwbNkypKWlYdWqVTA2NlbrWE3uQ1FREfLz81UWIiIiqqNEPZGfny8UCoX4/PPP1Sr/ySefCE9PT3k9IiJCtG/fXqM2AYiJEyeqbOvSpYuYNGmSEEKIzMxMAUCcPn1aCCHE/v37BQCRmJgol79x44ZQKpVi8+bNQgghgoKCxPjx41XqPHjwoGjQoIG4d++eEEKI5s2biyVLlqjEsWPHDpU216xZI+8/e/asACDS0tLUbuNJpKenCwBi7969le5ft26dMDMzU9m2Y8cO8fDHVN37EBERIQBUWOxDt4jmYd/KCxERET2/8vLyBACRl5dXbbl600OblpaGoqIi9OnTp9L9mzdvRrdu3WBjYwNjY2PMmzcP2dnZT92uj49PhfXH9dA+fIyFhQVcXV3lY1JTUxEfHw9jY2N58fPzQ1lZGTIzM9WOy93dXf7Z1tYWAJCbm6vVNh6VkpICPT099OzZ84nrUNecOXOQl5cnL5cvX67xNomIiKh21JuHwpRKZZX7jh49ioCAAERFRcHPzw9mZmZITExETEzMM4xQPQUFBZgwYQJCQkIq7HNwcFC7Hn19ffnn8jGqZWVlWm3jUdXdAwBo0KABhBAq28qHWmhKoVBAoVA80bFERESkW+pNQuvi4gKlUol9+/Zh7NixKvuOHDmC5s2bY+7cufK2S5cuqZQxMDBAaWmpxu0eO3YMgYGBKuseHh6PPaY8cbx16xbOnz8PNzc3AEDHjh1x7tw5ODs7axyLumqqjXbt2qGsrAwHDhxA3759K+y3srLCnTt38M8//8DIyAgAKszR+6T3gYiIiOquepPQNmrUCGFhYZg9ezYMDAzQrVs3/P333zh79ixcXFyQnZ2NxMREdOrUCUlJSdixY4fK8Y6OjsjMzERKSgqaNWsGExMTtXoAt27dCi8vL7z00kvYuHEjTpw4IT/VX5X58+ejcePGaNKkCebOnQtLS0sMGjQIwIPZGLy9vREcHIyxY8fCyMgI586dw969e7FixYonvj4Pq6k2HB0dMXLkSIwZMwbLli1D+/btcenSJeTm5mLo0KHo0qULDA0N8e9//xshISE4fvx4hVkMnvQ+EBERUd1Vb8bQAg9mHJgxYwbef/99uLm5YdiwYcjNzcXrr7+OadOmITg4GB06dMCRI0cQHh6ucuyQIUPQr18/9OrVC1ZWVkhISFCrzaioKCQmJsLd3R0bNmxAQkICWrduXe0xCxcuxNSpU+Hp6YmrV69i165dMDAwAPBg7OuBAwdw/vx5dO/eHR4eHnj//fdhZ2f3ZBelEjXZRlxcHN58801MnjwZrVq1wrhx4+TpwCwsLPDVV1/hu+++Q7t27ZCQkIDIyEiV45/0PhAREVHdJYlHBy2S1kiShB07dsi9q4+TnJyMXr164datWxXmY6Wnk5+fDzMzM9iHbkEDhaG8PWvhgFqMioiIiKpT/vs7Ly8PpqamVZarVz20RERERFT3MKF9Qhs3blSZ1urhpU2bNrUdXo05ePBgleetp6dX5T51X55AREREpKl681CYtr3++uvo0qVLpfvKp8TSdDSHr6+vxsc8a15eXhVmHih37969x07NRURERKRtHENL9YK6Y3CIiIjo+cExtERERERULzChJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaKleaRuxp7ZDICIiIi1jQktEREREOo0JLRERERHpNCa0RERERKTTmNASERERkU5jQqsljo6OiI2N1Wqdo0aNwqBBg+R1X19fhIaGarWNZy0yMhIdOnSotsyj501ERERUnYa1HUBdcfLkSRgZGdVoG9u3b4e+vn6NtqGuUaNG4fbt29i5c2dth0JERET1HBNaLbGysqrxNiwsLGq8DSIiIiJd89wNOfD19UVISAhmz54NCwsL2NjYIDIyEgCQlZUFSZKQkpIil799+zYkSUJycjIAIDk5GZIkYc+ePfDw8IBSqUTv3r2Rm5uL3bt3w83NDaampnj77bdx9+5dtWMKDg5GcHAwzMzMYGlpifDwcAgh5DKPDjmQJAlxcXHo378/lEolnJycsG3bNpV6L1++jKFDh8Lc3BwWFhbw9/dHVlZWtXE8POTA0dERH330EcaMGQMTExM4ODhg9erVT9WGOiIjI7F+/Xp8/fXXkCRJ5fqHhYWhZcuWMDQ0hJOTE8LDw1FSUlKhjlWrVsHe3h6GhoYYOnQo8vLyqmyvrKwM0dHRaNGiBZRKJdq3b1/hWhIREVH99dwltACwfv16GBkZ4fjx41i0aBHmz5+PvXv3alRHZGQkVqxYgSNHjshJXWxsLDZt2oSkpCT88MMPWL58uUYxNWzYECdOnMDSpUuxePFirFmzptpjwsPDMWTIEKSmpiIgIADDhw9HWloaAKCkpAR+fn4wMTHBwYMHcfjwYRgbG6Nfv34oLi5WO66YmBh4eXnh9OnTmDx5MiZNmoT09HSttvGomTNnYujQoejXrx9ycnKQk5ODrl27AgBMTEwQHx+Pc+fOYenSpfj888+xZMkSleMvXryILVu2YNeuXfj+++/l2KsSHR2NDRs24LPPPsPZs2cxbdo0vPPOOzhw4ECVxxQVFSE/P19lISIiojpKPGd69uwpXnrpJZVtnTp1EmFhYSIzM1MAEKdPn5b33bp1SwAQ+/fvF0IIsX//fgFA/Pjjj3KZ6OhoAUBkZGTI2yZMmCD8/PzUjsnNzU2UlZXJ28LCwoSbm5u83rx5c7FkyRJ5HYCYOHGiSj1dunQRkyZNEkII8eWXXwpXV1eVOouKioRSqRR79uwRQggxcuRI4e/vrxLH1KlTVdp855135PWysjJhbW0t4uLi1G7jST0aW1U++eQT4enpKa9HREQIPT09ceXKFXnb7t27RYMGDUROTk6FugsLC4WhoaE4cuSISr1BQUFixIgRVbYbEREhAFRY7EO3aHCWREREVJvy8vIEAJGXl1dtuedyDK27u7vKuq2tLXJzc5+4jiZNmsh/An9424kTJ9Suz9vbG5Ikyes+Pj6IiYlBaWkp9PT0Kj3Gx8enwnr5cInU1FRcvHgRJiYmKmUKCwuRkZGhdlwPn6ckSbCxsZGvlbba0MTmzZuxbNkyZGRkoKCgAPfv34epqalKGQcHBzRt2lRe9/HxQVlZGdLT02FjY6NS9uLFi7h79y5efvllle3FxcXw8PCoMo45c+Zg+vTp8np+fj7s7e2f5tSIiIjoOfVcJrSPPskvSRLKysrQoMGDERLiobGrlY3PfLQOSZKqrLO2FBQUwNPTExs3bqywT5MHzKo7L221oa6jR48iICAAUVFR8PPzg5mZGRITExETE/PEdRYUFAAAkpKSVJJgAFAoFFUep1Aoqt1PREREdcdzmdBWpTwJy8nJkXvnHn5ArCYdP35cZf3YsWNwcXGpsne2vExgYKDKenncHTt2xObNm2FtbV2hB1NbarINAwMDlJaWqmw7cuQImjdvjrlz58rbLl26VOHY7Oxs/PXXX7CzswPw4Lo0aNAArq6uFcq2bt0aCoUC2dnZ6Nmzp1bPgYiIiOqG5/KhsKoolUp4e3tj4cKFSEtLw4EDBzBv3rxn0nZ2djamT5+O9PR0JCQkYPny5Zg6dWq1x2zduhVr167F+fPnERERgRMnTiA4OBgAEBAQAEtLS/j7++PgwYPIzMxEcnIyQkJCcOXKFa3EXJNtODo64rfffkN6ejquX7+OkpISuLi4IDs7G4mJicjIyMCyZcuwY8eOCsc2atQII0eORGpqKg4ePIiQkBAMHTq0wnAD4MFDZjNnzsS0adOwfv16ZGRk4NSpU1i+fDnWr1//VOdAREREdYNOJbQAsHbtWty/fx+enp4IDQ3FggULnkm7gYGBuHfvHjp37owpU6Zg6tSpGD9+fLXHREVFITExEe7u7tiwYQMSEhLQunVrAIChoSF+/vlnODg4YPDgwXBzc0NQUBAKCwu11ptak22MGzcOrq6u8PLygpWVFQ4fPozXX38d06ZNQ3BwMDp06IAjR44gPDy8wrHOzs4YPHgwXn31Vbzyyitwd3fHypUrq2zrgw8+QHh4OKKjo+Hm5oZ+/fohKSkJLVq0eKpzICIiorpBEg8PSKVK+fr6okOHDhq92laSJOzYsYOvcH1O5Ofnw8zMDPahW5C95K3aDoeIiIjUUP77Oy8vr9rOOJ3roSUiIiIiephOPRRWE7Kzs+VhAJU5d+7cM4zm2TI2Nq5y371796BUKivdt3v3bnTv3r2mwiIiIiLSSL1PaO3s7KqdKcHOzk5+rasmdGEkR3XnXV1C++j0WURERES1iWNoqV5QdwwOERERPT84hpaIiIiI6gUmtERERESk05jQEhEREZFOY0JLRERERDqNCS0RERER6TQmtERERESk05jQEhEREZFOY0JLRERERDqNCS0RERER6TQmtERERESk05jQEhEREZFOY0JLRERERDqNCS0RERER6TQmtERERESk05jQEhEREZFOY0JLRERERDqNCS0RERER6TQmtERERESk05jQEhEREZFOY0JLRERERDqNCS3VK20j9tR2CERERKRlTGiJiIiISKcxoSUiIiIincaEloiIiIh0GhNaIiIiItJpTGifgKOjI2JjY7Va56hRozBo0CB53dfXF6GhoVptQ5sejbcm1cT1JiIiorqjYW0HoItOnjwJIyOjGm1j+/bt0NfXr9E2njfx8fEIDQ3F7du3azsUIiIi0iFMaJ+AlZVVjbdhYWFR420QERER1QW1OuTA19cXISEhmD17NiwsLGBjY4PIyEgAQFZWFiRJQkpKilz+9u3bkCQJycnJAIDk5GRIkoQ9e/bAw8MDSqUSvXv3Rm5uLnbv3g03NzeYmpri7bffxt27d9WOKTg4GMHBwTAzM4OlpSXCw8MhhJDLPPoncEmSEBcXh/79+0OpVMLJyQnbtm1Tqffy5csYOnQozM3NYWFhAX9/f2RlZVUbx8NDDhwdHfHRRx9hzJgxMDExgYODA1avXv1UbairtLQU06dPh7m5ORo3bozZs2erXA8AKCsrQ3R0NFq0aAGlUon27durXIPye5WUlAR3d3c0atQI3t7eOHPmjLx/9OjRyMvLgyRJkCRJ/iwAwN27d6s9dyIiIqq/an0M7fr162FkZITjx49j0aJFmD9/Pvbu3atRHZGRkVixYgWOHDkiJ3WxsbHYtGkTkpKS8MMPP2D58uUaxdSwYUOcOHECS5cuxeLFi7FmzZpqjwkPD8eQIUOQmpqKgIAADB8+HGlpaQCAkpIS+Pn5wcTEBAcPHsThw4dhbGyMfv36obi4WO24YmJi4OXlhdOnT2Py5MmYNGkS0tPTtdpGVe3Gx8dj7dq1OHToEG7evIkdO3aolImOjsaGDRvw2Wef4ezZs5g2bRreeecdHDhwQKXcrFmzEBMTg5MnT8LKygoDBw5ESUkJunbtitjYWJiamiInJwc5OTmYOXOmWudemaKiIuTn56ssREREVEeJWtSzZ0/x0ksvqWzr1KmTCAsLE5mZmQKAOH36tLzv1q1bAoDYv3+/EEKI/fv3CwDixx9/lMtER0cLACIjI0PeNmHCBOHn56d2TG5ubqKsrEzeFhYWJtzc3OT15s2biyVLlsjrAMTEiRNV6unSpYuYNGmSEEKIL7/8Uri6uqrUWVRUJJRKpdizZ48QQoiRI0cKf39/lTimTp2q0uY777wjr5eVlQlra2sRFxendhtPytbWVixatEheLykpEc2aNZPjLSwsFIaGhuLIkSMqxwUFBYkRI0YIIf7vXiUmJsr7b9y4IZRKpdi8ebMQQoh169YJMzOzCu0/7twrExERIQBUWOxDt2h8/kRERFQ78vLyBACRl5dXbbla76F1d3dXWbe1tUVubu4T19GkSRMYGhrCyclJZZsmdXp7e0OSJHndx8cHFy5cQGlpaZXH+Pj4VFgv76FNTU3FxYsXYWJiAmNjYxgbG8PCwgKFhYXIyMhQO66Hz1OSJNjY2Mjnpa02HpWXl4ecnBx06dJF3tawYUN4eXnJ6xcvXsTdu3fx8ssvy20bGxtjw4YNFdp++DpZWFjA1dVVvk5Peu6VmTNnDvLy8uTl8uXLap0vERER6Z5afyjs0Sf5JUlCWVkZGjR4kGuLh8ZqlpSUPLYOSZKqrLO2FBQUwNPTExs3bqywT5MHzKo7L2218SQKCgoAAElJSWjatKnKPoVCoZU2NL2nCoVCa20TERHR863WE9qqlCdhOTk58PDwAACVB8Rq0vHjx1XWjx07BhcXF+jp6VV5zLFjxxAYGKiyXh53x44dsXnzZlhbW8PU1LRGYq6pNszMzGBra4vjx4+jR48eAID79+/j119/RceOHQEArVu3hkKhQHZ2Nnr27FltfceOHYODgwMA4NatWzh//jzc3NwAAAYGBtX2ghMRERFVptaHHFRFqVTC29sbCxcuRFpaGg4cOIB58+Y9k7azs7Mxffp0pKenIyEhAcuXL8fUqVOrPWbr1q1Yu3Ytzp8/j4iICJw4cQLBwcEAgICAAFhaWsLf3x8HDx5EZmYmkpOTERISgitXrmgl5ppsY+rUqVi4cCF27tyJ33//HZMnT1aZK9bExAQzZ87EtGnTsH79emRkZODUqVNYvnw51q9fr1LX/PnzsW/fPpw5cwajRo2CpaWl/IIGR0dHFBQUYN++fbh+/braM1MQERFR/fbcJrQAsHbtWty/fx+enp4IDQ3FggULnkm7gYGBuHfvHjp37owpU6Zg6tSpGD9+fLXHREVFITExEe7u7tiwYQMSEhLQunVrAIChoSF+/vlnODg4YPDgwXBzc0NQUBAKCwu11ptak23MmDED7777LkaOHAkfHx+YmJjgjTfeUCnzwQcfIDw8HNHR0XBzc0O/fv2QlJSEFi1aqJRbuHAhpk6dCk9PT1y9ehW7du2CgYEBAKBr166YOHEihg0bBisrKyxatOip4iYiIqL6QRLikQlF6zlfX1906NBBo1etSpKEHTt2PLNXweqi5ORk9OrVC7du3YK5ufkzbz8/Px9mZmawD92C7CVvPfP2iYiISHPlv7/z8vKq7aB7rntoiYiIiIge57l9KKwmZGdny8MAKnPu3LlnGM2zZWxsXOW+e/fuQalUVrpv9+7d6N69e02FRURERPTU6lVCa2dnV+1MCXZ2dvJrdTWhC6M2qjvv6hLaR6fhelK+vr46cZ2IiIhI93AMLdUL6o7BISIioucHx9ASERERUb3AhJaIiIiIdBoTWiIiIiLSaUxoiYiIiEinMaElIiIiIp3GhJaIiIiIdBoTWiIiIiLSaUxoiYiIiEinMaElIiIiIp1Wr159S/VX+Qvx8vPzazkSIiIiUlf57+3HvdiWCS3VCzdu3AAA2Nvb13IkREREpKk7d+7AzMysyv1MaKlesLCwAABkZ2dX+4Wg50d+fj7s7e1x+fLlat/fTc8H3i/dw3ume+rjPRNC4M6dO7Czs6u2HBNaqhcaNHgwXNzMzKze/CNQV5iamvKe6RDeL93De6Z76ts9U6cjig+FEREREZFOY0JLRERERDqNCS3VCwqFAhEREVAoFLUdCqmJ90y38H7pHt4z3cN7VjVJPG4eBCIiIiKi5xh7aImIiIhIpzGhJSIiIiKdxoSWiIiIiHQaE1oiIiIi0mlMaKnO+/TTT+Ho6IhGjRqhS5cuOHHiRG2HRFWIjIyEJEkqS6tWrWo7LHrIzz//jIEDB8LOzg6SJGHnzp0q+4UQeP/992FrawulUom+ffviwoULtRMsAXj8PRs1alSF712/fv1qJ1hCdHQ0OnXqBBMTE1hbW2PQoEFIT09XKVNYWIgpU6agcePGMDY2xpAhQ3Dt2rVaivj5wISW6rTNmzdj+vTpiIiIwKlTp9C+fXv4+fkhNze3tkOjKrRp0wY5OTnycujQodoOiR7yzz//oH379vj0008r3b9o0SIsW7YMn332GY4fPw4jIyP4+fmhsLDwGUdK5R53zwCgX79+Kt+7hISEZxghPezAgQOYMmUKjh07hr1796KkpASvvPIK/vnnH7nMtGnTsGvXLmzduhUHDhzAX3/9hcGDB9di1M8BQVSHde7cWUyZMkVeLy0tFXZ2diI6OroWo6KqREREiPbt29d2GKQmAGLHjh3yellZmbCxsRGffPKJvO327dtCoVCIhISEWoiQHvXoPRNCiJEjRwp/f/9aiYceLzc3VwAQBw4cEEI8+E7p6+uLrVu3ymXS0tIEAHH06NHaCrPWsYeW6qzi4mL8+uuv6Nu3r7ytQYMG6Nu3L44ePVqLkVF1Lly4ADs7Ozg5OSEgIADZ2dm1HRKpKTMzE1evXlX5zpmZmaFLly78zj3nkpOTYW1tDVdXV0yaNAk3btyo7ZDo/8vLywMAWFhYAAB+/fVXlJSUqHzPWrVqBQcHh3r9PWNCS3XW9evXUVpaiiZNmqhsb9KkCa5evVpLUVF1unTpgvj4eHz//feIi4tDZmYmunfvjjt37tR2aKSG8u8Vv3O6pV+/ftiwYQP27duHjz/+GAcOHED//v1RWlpa26HVe2VlZQgNDUW3bt3Qtm1bAA++ZwYGBjA3N1cpW9+/Zw1rOwAionL9+/eXf3Z3d0eXLl3QvHlzbNmyBUFBQbUYGVHdNXz4cPnndu3awd3dHS+++CKSk5PRp0+fWoyMpkyZgjNnzvBZAjWwh5bqLEtLS+jp6VV48vPatWuwsbGppahIE+bm5mjZsiUuXrxY26GQGsq/V/zO6TYnJydYWlrye1fLgoOD8e2332L//v1o1qyZvN3GxgbFxcW4ffu2Svn6/j1jQkt1loGBATw9PbFv3z55W1lZGfbt2wcfH59ajIzUVVBQgIyMDNja2tZ2KKSGFi1awMbGRuU7l5+fj+PHj/M7p0OuXLmCGzdu8HtXS4QQCA4Oxo4dO/DTTz+hRYsWKvs9PT2hr6+v8j1LT09HdnZ2vf6eccgB1WnTp0/HyJEj4eXlhc6dOyM2Nhb//PMPRo8eXduhUSVmzpyJgQMHonnz5vjrr78QEREBPT09jBgxorZDo/+voKBApecuMzMTKSkpsLCwgIODA0JDQ7FgwQK4uLigRYsWCA8Ph52dHQYNGlR7Qddz1d0zCwsLREVFYciQIbCxsUFGRgZmz54NZ2dn+Pn51WLU9deUKVOwadMmfP311zAxMZHHxZqZmUGpVMLMzAxBQUGYPn06LCwsYGpqivfeew8+Pj7w9vau5ehrUW1Ps0BU05YvXy4cHByEgYGB6Ny5szh27Fhth0RVGDZsmLC1tRUGBgaiadOmYtiwYeLixYu1HRY9ZP/+/QJAhWXkyJFCiAdTd4WHh4smTZoIhUIh+vTpI9LT02s36Hquunt29+5d8corrwgrKyuhr68vmjdvLsaNGyeuXr1a22HXW5XdKwBi3bp1cpl79+6JyZMnixdeeEEYGhqKN954Q+Tk5NRe0M8BSQghnn0aTURERESkHRxDS0REREQ6jQktEREREek0JrREREREpNOY0BIRERGRTmNCS0REREQ6jQktEREREek0JrREREREpNOY0BIR6TAhBMaPHw8LCwtIkoSUlJTaDklrJEnCzp07azsMItIBTGiJiHTY999/j/j4eHz77bfIyclB27ZttVLvqFGjav11tTk5Oejfv3+txlCdyMhIdOjQobbDICIADWs7ACIienIZGRmwtbVF165dazuUSpWWlkKSJDRooHn/iY2NTQ1E9PSEECgtLa3tMIjoIeyhJSLSUaNGjcJ7772H7OxsSJIER0dHAEBZWRmio6PRokULKJVKtG/fHtu2bZOPKy0tRVBQkLzf1dUVS5culfdHRkZi/fr1+PrrryFJEiRJQnJyMpKTkyFJEm7fvi2XTUlJgSRJyMrKAgDEx8fD3Nwc33zzDVq3bg2FQoHs7GwUFRVh5syZaNq0KYyMjNClSxckJydXe34PDznIysqCJEnYsmULunfvDqVSiU6dOuH8+fM4efIkvLy8YGxsjP79++Pvv/9WuUaDBg1CVFQUrKysYGpqiokTJ6K4uFguU1RUhJCQEFhbW6NRo0Z46aWXcPLkSXl/+Xnv3r0bnp6eUCgU+OqrrxAVFYXU1FT5GsXHxwMAFi9ejHbt2sHIyAj29vaYPHkyCgoK5PrKr9GePXvg5uYGY2Nj9OvXDzk5OSrnv3btWrRp0wYKhQK2trYIDg6W992+fRtjx46Vz6l3795ITU2t9noS1WmCiIh00u3bt8X8+fNFs2bNRE5OjsjNzRVCCLFgwQLRqlUr8f3334uMjAyxbt06oVAoRHJyshBCiOLiYvH++++LkydPij/++EN89dVXwtDQUGzevFkIIcSdO3fE0KFDRb9+/UROTo7IyckRRUVFYv/+/QKAuHXrlhzD6dOnBQCRmZkphBBi3bp1Ql9fX3Tt2lUcPnxY/P777+Kff/4RY8eOFV27dhU///yzuHjxovjkk0+EQqEQ58+fr/L8AIgdO3YIIYTIzMwUAOTzOnfunPD29haenp7C19dXHDp0SJw6dUo4OzuLiRMnynWMHDlSGBsbi2HDhokzZ86Ib7/9VlhZWYl///vfcpmQkBBhZ2cnvvvuO3H27FkxcuRI8cILL4gbN24IIYR83u7u7uKHH34QFy9eFFeuXBEzZswQbdq0ka/R3bt3hRBCLFmyRPz0008iMzNT7Nu3T7i6uopJkybJ7ZVfo759+4qTJ0+KX3/9Vbi5uYm3335bLrNy5UrRqFEjERsbK9LT08WJEyfEkiVL5P19+/YVAwcOFCdPnhTnz58XM2bMEI0bN5ZjJqpvmNASEemwJUuWiObNm8vrhYWFwtDQUBw5ckSlXFBQkBgxYkSV9UyZMkUMGTJEXh85cqTw9/dXKaNuQgtApKSkyGUuXbok9PT0xJ9//qlSX58+fcScOXOqjKmyhHbNmjXy/oSEBAFA7Nu3T94WHR0tXF1dVc7DwsJC/PPPP/K2uLg4YWxsLEpLS0VBQYHQ19cXGzdulPcXFxcLOzs7sWjRIpXz3rlzp0p8ERERon379lXGX27r1q2icePG8nr5Nbp48aK87dNPPxVNmjSR1+3s7MTcuXMrre/gwYPC1NRUFBYWqmx/8cUXxapVqx4bD1FdxDG0RER1yMWLF3H37l28/PLLKtuLi4vh4eEhr3/66adYu3YtsrOzce/ePRQXF2vtAScDAwO4u7vL6//73/9QWlqKli1bqpQrKipC48aNNar74XqbNGkCAGjXrp3KttzcXJVj2rdvD0NDQ3ndx8cHBQUFuHz5MvLy8lBSUoJu3brJ+/X19dG5c2ekpaWp1OPl5aVWjD/++COio6Px+++/Iz8/H/fv30dhYSHu3r0rx2FoaIgXX3xRPsbW1laOOzc3F3/99Rf69OlTaf2pqakoKCiocO3u3buHjIwMtWIkqmuY0BIR1SHlYzWTkpLQtGlTlX0KhQIAkJiYiJkzZyImJgY+Pj4wMTHBJ598guPHj1dbd/mDXUIIeVtJSUmFckqlEpIkqcSkp6eHX3/9FXp6eipljY2NNTi7B8lmufI2Ht1WVlamUZ3qMjIyemyZrKwsvPbaa5g0aRI+/PBDWFhY4NChQwgKCkJxcbGc0D4cc3nc5ddVqVRW20ZBQQFsbW0rHYNsbm6u3skQ1TFMaImI6pCHH8Tq2bNnpWUOHz6Mrl27YvLkyfK2R3v2DAwMKjzJb2VlBeDBdFovvPACAKg1762HhwdKS0uRm5uL7t27a3I6WpGamop79+7JieKxY8dgbGwMe3t7WFpawsDAAIcPH0bz5s0BPEjST548idDQ0Grrrewa/frrrygrK0NMTIz8H4AtW7ZoFK+JiQkcHR2xb98+9OrVq8L+jh074urVq2jYsKH8ICBRfcdZDoiI6hATExPMnDkT06ZNw/r165GRkYFTp05h+fLlWL9+PQDAxcUFv/zyC/bs2YPz588jPDxc5al+AHB0dMRvv/2G9PR0XL9+HSUlJXB2doa9vT0iIyNx4cIFJCUlISYm5rExtWzZEgEBAQgMDMT27duRmZmJEydOIDo6GklJSTVyHR5WXFyMoKAgnDt3Dt999x0iIiIQHByMBg0awMjICJMmTcKsWbPw/fff49y5cxg3bhzu3r2LoKCgaut1dHREZmYmUlJScP36dRQVFcHZ2RklJSVYvnw5/vjjD3z55Zf47LPPNI45MjISMTExWLZsGS5cuCDfQwDo27cvfHx8MGjQIPzwww/IysrCkSNHMHfuXPzyyy9PdI2IdB0TWiKiOuaDDz5AeHg4oqOj4ebmhn79+iEpKQktWrQAAEyYMAGDBw/GsGHD0KVLF9y4cUOltxYAxo0bB1dXV3h5ecHKygqHDx+Gvr4+EhIS8Pvvv8Pd3R0ff/wxFixYoFZM69atQ2BgIGbMmAFXV1cMGjQIJ0+ehIODg9bP/1F9+vSBi4sLevTogWHDhuH1119HZGSkvH/hwoUYMmQI3n33XXTs2BEXL17Enj175F7oqgwZMgT9+vVDr169YGVlhYSEBLRv3x6LFy/Gxx9/jLZt22Ljxo2Ijo7WOOaRI0ciNjYWK1euRJs2bfDaa6/hwoULAB4MT/juu+/Qo0cPjB49Gi1btsTw4cNx6dIleVwxUX0jiYcHQxEREdUho0aNwu3bt/kKXaI6jj20RERERKTTmNASERERkU7jkAMiIiIi0mnsoSUiIiIincaEloiIiIh0GhNaIiIiItJpTGiJiIiISKcxoSUiIiIincaEloiIiIh0GhNaIiIiItJpTGiJiIiISKcxoSUiIiIinfb/AFODkLlOeJJ2AAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "feature_imp = best_cbr.feature_importances_\n",
    "feature_nm = best_cbr.feature_names_\n",
    "imp_series = pd.Series(feature_imp)\n",
    "imp_series.index = feature_nm\n",
    "print(imp_series.sort_values(ascending=False))\n",
    "print('\\n')\n",
    "imp_series.sort_values().plot(kind='barh',\n",
    "                              xlabel='feature importance',\n",
    "                              ylabel='feature name',\n",
    "                              title='Catboost Feature importances')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "venv",
   "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": "3.10.7"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}