classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Import"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "import re\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from pycaret.datasets import get_data\n",
    "from lightgbm import LGBMClassifier\n",
    "from sklearn.metrics import log_loss, average_precision_score\n",
    "from sklearn.dummy import DummyClassifier\n",
    "from tqdm import tqdm\n",
    "from shap import TreeExplainer\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Utils"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def shap_sum2proba(shap_sum):\n",
    "  \"\"\"Compute sigmoid function of the Shap sum to get predicted probability.\"\"\"\n",
    "  \n",
    "  return 1 / (1 + np.exp(-shap_sum))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def individual_log_loss(y_true, y_pred, eps = 1e-15):\n",
    "  \"\"\"Compute log-loss for each individual of the sample.\"\"\"\n",
    "  \n",
    "  y_pred = np.clip(y_pred, eps, 1 - eps)\n",
    "  return - y_true * np.log(y_pred) - (1 - y_true) * np.log(1 - y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_preds_shaps(df, features, target, ix_trn):\n",
    "  \"\"\"Get predictions (predicted probabilities) and SHAP values for a dataset.\"\"\"\n",
    "    \n",
    "  model = LGBMClassifier().fit(df.loc[ix_trn, features], df.loc[ix_trn, target])\n",
    "  preds = pd.Series(model.predict_proba(df[features])[:,1], index=df.index)\n",
    "  shap_explainer = TreeExplainer(model)\n",
    "  shap_expected_value = shap_explainer.expected_value[-1]\n",
    "  shaps = pd.DataFrame(\n",
    "    data=shap_explainer.shap_values(df[features])[1],\n",
    "    index=df.index,\n",
    "    columns=features)\n",
    "  return preds, shaps, shap_expected_value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_feature_contributions(y_true, y_pred, shap_values, shap_expected_value):\n",
    "  \"\"\"Compute prediction contribution and error contribution for each feature.\"\"\"\n",
    "\n",
    "  prediction_contribution = shap_values.abs().mean().rename(\"prediction_contribution\")\n",
    "  \n",
    "  ind_log_loss = individual_log_loss(y_true=y_true, y_pred=y_pred).rename(\"log_loss\")\n",
    "  y_pred_wo_feature = shap_values.apply(lambda feature: shap_expected_value + shap_values.sum(axis=1) - feature).applymap(shap_sum2proba)\n",
    "  ind_log_loss_wo_feature = y_pred_wo_feature.apply(lambda feature: individual_log_loss(y_true=y_true, y_pred=feature))\n",
    "  ind_log_loss_diff = ind_log_loss_wo_feature.apply(lambda feature: ind_log_loss - feature)\n",
    "  error_contribution = ind_log_loss_diff.mean().rename(\"error_contribution\").T\n",
    "  \n",
    "  return prediction_contribution, error_contribution"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4. Toy Example"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.1 Generate data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "shap_values = pd.DataFrame([\n",
    "  [-2.0, 0.1, 0.8],\n",
    "  [3.0,  0., 0.5],\n",
    "  [-5.0, -2.0, 3.0]], columns=[\"job\", \"age\", \"nationality\"])\n",
    "y_pred = shap_values.sum(axis=1).apply(shap_sum2proba).rename(\"income>100k (predicted)\")\n",
    "y_true = pd.Series([0, 1, 0], name=\"income>100k (true)\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.2 Compute Prediction Contribution and Error Contribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "prediction_contribution = shap_values.abs().mean().rename(\"prediction_contribution\").to_frame().T\n",
    "\n",
    "ind_log_loss = individual_log_loss(y_true=y_true, y_pred=y_pred).rename(\"log_loss\")\n",
    "y_pred_wo_feature = shap_values.apply(lambda feature: shap_values.sum(axis=1) - feature).applymap(shap_sum2proba)\n",
    "ind_log_loss_wo_feature = y_pred_wo_feature.apply(lambda feature: individual_log_loss(y_true=y_true, y_pred=feature))\n",
    "ind_log_loss_diff = ind_log_loss_wo_feature.apply(lambda feature: ind_log_loss - feature)\n",
    "error_contribution = ind_log_loss_diff.mean().rename(\"error_contribution\").to_frame().T"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.3 Display items"
   ]
  },
  {
   "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>income&gt;100k (true)</th>\n",
       "      <th>income&gt;100k (predicted)</th>\n",
       "      <th>log_loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0.249740</td>\n",
       "      <td>0.287335</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>0.970688</td>\n",
       "      <td>0.029750</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>0.017986</td>\n",
       "      <td>0.018150</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   income>100k (true)  income>100k (predicted)  log_loss\n",
       "0                   0                 0.249740  0.287335\n",
       "1                   1                 0.970688  0.029750\n",
       "2                   0                 0.017986  0.018150"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat([y_true, y_pred, ind_log_loss], axis=1)"
   ]
  },
  {
   "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>job</th>\n",
       "      <th>age</th>\n",
       "      <th>nationality</th>\n",
       "      <th>shap sum</th>\n",
       "      <th>predicted probability</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-2.0</td>\n",
       "      <td>0.1</td>\n",
       "      <td>0.8</td>\n",
       "      <td>-1.1</td>\n",
       "      <td>0.249740</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>3.5</td>\n",
       "      <td>0.970688</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-5.0</td>\n",
       "      <td>-2.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>-4.0</td>\n",
       "      <td>0.017986</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   job  age  nationality  shap sum  predicted probability\n",
       "0 -2.0  0.1          0.8      -1.1               0.249740\n",
       "1  3.0  0.0          0.5       3.5               0.970688\n",
       "2 -5.0 -2.0          3.0      -4.0               0.017986"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat([shap_values, shap_values.sum(axis=1).rename(\"shap sum\"), y_pred.rename(\"predicted probability\")], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>job</th>\n",
       "      <th>age</th>\n",
       "      <th>nationality</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>prediction_contribution</th>\n",
       "      <td>3.333333</td>\n",
       "      <td>0.7</td>\n",
       "      <td>1.433333</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                              job  age  nationality\n",
       "prediction_contribution  3.333333  0.7     1.433333"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prediction_contribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>job</th>\n",
       "      <th>age</th>\n",
       "      <th>nationality</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.710950</td>\n",
       "      <td>0.231475</td>\n",
       "      <td>0.130108</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.622459</td>\n",
       "      <td>0.970688</td>\n",
       "      <td>0.952574</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.731059</td>\n",
       "      <td>0.119203</td>\n",
       "      <td>0.000911</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        job       age  nationality\n",
       "0  0.710950  0.231475     0.130108\n",
       "1  0.622459  0.970688     0.952574\n",
       "2  0.731059  0.119203     0.000911"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_wo_feature"
   ]
  },
  {
   "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>job</th>\n",
       "      <th>age</th>\n",
       "      <th>nationality</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1.241154</td>\n",
       "      <td>0.263282</td>\n",
       "      <td>0.139387</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.474077</td>\n",
       "      <td>0.029750</td>\n",
       "      <td>0.048587</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1.313262</td>\n",
       "      <td>0.126928</td>\n",
       "      <td>0.000911</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        job       age  nationality\n",
       "0  1.241154  0.263282     0.139387\n",
       "1  0.474077  0.029750     0.048587\n",
       "2  1.313262  0.126928     0.000911"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ind_log_loss_wo_feature"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>job</th>\n",
       "      <th>age</th>\n",
       "      <th>nationality</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.953819</td>\n",
       "      <td>0.024053</td>\n",
       "      <td>0.147949</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-0.444327</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.018837</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-1.295112</td>\n",
       "      <td>-0.108778</td>\n",
       "      <td>0.017238</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        job       age  nationality\n",
       "0 -0.953819  0.024053     0.147949\n",
       "1 -0.444327  0.000000    -0.018837\n",
       "2 -1.295112 -0.108778     0.017238"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ind_log_loss_diff"
   ]
  },
  {
   "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>job</th>\n",
       "      <th>age</th>\n",
       "      <th>nationality</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>error_contribution</th>\n",
       "      <td>-0.897752</td>\n",
       "      <td>-0.028242</td>\n",
       "      <td>0.048783</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         job       age  nationality\n",
       "error_contribution -0.897752 -0.028242     0.048783"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "error_contribution"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. Theory"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$$\n",
    "\\frac{1}{N} \\sum_{i=1}^{N} - y_i \\cdot \\log ( p_i ) - ( 1 - y_i ) \\cdot \\log ( 1 - p_i ) \n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$$\n",
    "- y \\cdot \\log(p) - ( 1 - y ) \\cdot \\log ( 1 - p ) \n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$$\n",
    "\\begin{cases}\n",
    "-\\log(p) & \\text{if $y$ = 1}\\\\\n",
    "-\\log(1-p) & \\text{if $y$ = 0}\\\\\n",
    "\\end{cases} \n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "\n",
    "p = np.linspace(.0001, .9999, 1000)\n",
    "\n",
    "ax.plot(p, -np.log(p), lw=3, label = \"if y=1 → -log(p)\")\n",
    "ax.plot(p, -np.log(1-p), lw=3, label = \"if y=0 → -log(1-p)\")\n",
    "ax.set_xlabel(\"p\", fontsize=12)\n",
    "ax.set_ylabel(\"individual log loss\", fontsize=12)\n",
    "ax.legend(fontsize=12)\n",
    "ax.grid()\n",
    "ax.tick_params(axis=\"both\", which=\"major\", labelsize=12)\n",
    "\n",
    "fig.savefig(\"logloss_curves_.png\", dpi=200, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$$\n",
    "p=\\frac{1}{1 + e^{\\,-\\,\\mathrm{shap\\,sum}}}\n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "\n",
    "shap_sum = np.linspace(-10, 10, 1000)\n",
    "\n",
    "ax.plot(shap_sum, shap_sum2proba(shap_sum), lw=3, label = \"1 / (1 + exp(-x))\")\n",
    "\n",
    "for i in range(3):\n",
    "  ax.scatter(shap_values.sum(axis=1).loc[i], y_pred.loc[i], color=\"red\", edgecolors=\"black\", s=80, zorder=100)\n",
    "\n",
    "ax.set_xlabel(\"shap sum\", fontsize=12)\n",
    "ax.set_ylabel(\"predicted probability\", fontsize=12)\n",
    "ax.legend(fontsize=12)\n",
    "ax.grid()\n",
    "ax.tick_params(axis=\"both\", which=\"major\", labelsize=12)\n",
    "\n",
    "fig.savefig(\"sigmoid_func_.png\", dpi=200, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5. Real dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "#df = get_data(\"gold\", verbose=False)\n",
    "#df *= 100\n",
    "#df[\"Gold_T+22\"] = (df[\"Gold_T+22\"] > 5).astype(int)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"gold.csv\", sep=\"\\t\")\n",
    "\n",
    "target = \"Gold_T+22\"\n",
    "\n",
    "features = [f for f in df.columns if f != target]\n",
    "argsort = np.argsort([f.split(\"T-\")[0] + \"T-\" + \"{:2.1f}\".format(float(f.split(\"T-\")[1])).zfill(4) for f in features])\n",
    "features = [features[arg] for arg in argsort]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 6. Data Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    2558.000000\n",
       "mean        0.145035\n",
       "std         0.352205\n",
       "min         0.000000\n",
       "25%         0.000000\n",
       "50%         0.000000\n",
       "75%         0.000000\n",
       "max         1.000000\n",
       "Name: Gold_T+22, dtype: float64"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[target].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "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>3M Libor_T-1</th>\n",
       "      <th>3M Libor_T-7</th>\n",
       "      <th>3M Libor_T-14</th>\n",
       "      <th>3M Libor_T-22</th>\n",
       "      <th>...</th>\n",
       "      <th>Yuan_T-7</th>\n",
       "      <th>Yuan_T-14</th>\n",
       "      <th>Yuan_T-22</th>\n",
       "      <th>Gold_T+22</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.00</td>\n",
       "      <td>-1.80</td>\n",
       "      <td>-2.90</td>\n",
       "      <td>-4.07</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.01</td>\n",
       "      <td>-0.02</td>\n",
       "      <td>-0.00</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-0.46</td>\n",
       "      <td>-1.47</td>\n",
       "      <td>-3.34</td>\n",
       "      <td>-4.51</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.01</td>\n",
       "      <td>-0.04</td>\n",
       "      <td>-0.01</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2556</th>\n",
       "      <td>-0.23</td>\n",
       "      <td>0.02</td>\n",
       "      <td>-1.18</td>\n",
       "      <td>-6.73</td>\n",
       "      <td>...</td>\n",
       "      <td>1.25</td>\n",
       "      <td>1.88</td>\n",
       "      <td>3.98</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2557</th>\n",
       "      <td>-0.29</td>\n",
       "      <td>-0.83</td>\n",
       "      <td>-1.93</td>\n",
       "      <td>-5.03</td>\n",
       "      <td>...</td>\n",
       "      <td>1.21</td>\n",
       "      <td>2.22</td>\n",
       "      <td>3.46</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2558 rows × 121 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      3M Libor_T-1  3M Libor_T-7  3M Libor_T-14  3M Libor_T-22  ...  Yuan_T-7  \\\n",
       "0             0.00         -1.80          -2.90          -4.07  ...     -0.01   \n",
       "1            -0.46         -1.47          -3.34          -4.51  ...     -0.01   \n",
       "...            ...           ...            ...            ...  ...       ...   \n",
       "2556         -0.23          0.02          -1.18          -6.73  ...      1.25   \n",
       "2557         -0.29         -0.83          -1.93          -5.03  ...      1.21   \n",
       "\n",
       "      Yuan_T-14  Yuan_T-22  Gold_T+22  \n",
       "0         -0.02      -0.00          0  \n",
       "1         -0.04      -0.01          0  \n",
       "...         ...        ...        ...  \n",
       "2556       1.88       3.98          0  \n",
       "2557       2.22       3.46          0  \n",
       "\n",
       "[2558 rows x 121 columns]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.set_option('display.max_columns', 8)\n",
    "pd.set_option(\"display.max_rows\", 5)\n",
    "\n",
    "df[features + [target]].round(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3M Libor            AUD                 Brent 1             Brent 2             \n",
      "Corn                Crude Palm Oil      DJ Index            Dollex              \n",
      "EM Bond             EMEquity ETF        Euro Curr           Gold                \n",
      "INR                 Light Crude         MR                  MSCI APAC ex J      \n",
      "MSCI EM             Nifty               Nikkei Ind          Palladium           \n",
      "Platinum            S&P 500             Shanghai Index      Silver              \n",
      "US Bond ETF         Wheat 1             Wheat 2             White Sugar         \n",
      "Yuan                "
     ]
    }
   ],
   "source": [
    "for enum, f in enumerate(sorted(set([f.split(\"_\")[0] for f in features]))):\n",
    "  print(f.ljust(20), end=\"\")\n",
    "  if enum % 4 == 3:\n",
    "    print(\"\\n\", end=\"\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 7. Train/Validation/Test Split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(123)\n",
    "fold = np.random.choice([\"trn\", \"val\", \"tst\"], p=[1/3, 1/3, 1/3], size=len(df))\n",
    "\n",
    "ix_trn = fold == \"trn\"\n",
    "ix_val = fold == \"val\"\n",
    "ix_tst = fold == \"tst\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "assert (ix_trn.astype(int) + ix_val.astype(int) + ix_tst.astype(int) == 1).all()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 8. Single Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.1 Compute Predictions and Shap Values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds, shaps, shap_expected_value = get_preds_shaps(df=df, features=features, target=target, ix_trn=ix_trn)\n",
    "\n",
    "assert ((preds - (shap_expected_value + shaps.sum(axis=1)).apply(shap_sum2proba)).abs() < 1e-10).all()\n",
    "\n",
    "prediction_contribution_trn, error_contribution_trn = get_feature_contributions(\n",
    "  y_true=df.loc[ix_trn, target], \n",
    "  y_pred=preds.loc[ix_trn], \n",
    "  shap_values=shaps.loc[ix_trn, :],\n",
    "  shap_expected_value=shap_expected_value\n",
    ")\n",
    "\n",
    "prediction_contribution_val, error_contribution_val = get_feature_contributions(\n",
    "  y_true=df.loc[ix_val, target], \n",
    "  y_pred=preds.loc[ix_val], \n",
    "  shap_values=shaps.loc[ix_val, :],\n",
    "  shap_expected_value=shap_expected_value\n",
    ")\n",
    "\n",
    "contributions_trn = pd.concat([prediction_contribution_trn, error_contribution_trn], axis=1)\n",
    "contributions_val = pd.concat([prediction_contribution_val, error_contribution_val], axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.2 Write/Read Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "contributions_trn.to_csv(\"contributions_trn.csv\", index=True)\n",
    "contributions_val.to_csv(\"contributions_val.csv\", index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "contributions_trn = pd.read_csv(\"contributions_trn.csv\", index_col=0)\n",
    "contributions_val = pd.read_csv(\"contributions_val.csv\", index_col=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.3 Prediction vs. Error Contribution (Training)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "\n",
    "ax.scatter(contributions_trn[\"prediction_contribution\"], contributions_trn[\"error_contribution\"], s=80, color=\"orange\", edgecolors=\"black\")\n",
    "ax.hlines(y=0, xmin=0, xmax=contributions_trn[\"prediction_contribution\"].max(), ls=\"--\", color=\"black\")\n",
    "ax.tick_params(axis=\"both\", which=\"major\", labelsize=12)\n",
    "ax.grid(); ax.set_axisbelow(True)\n",
    "ax.set_title(\"Train Dataset\", fontsize=12)\n",
    "ax.set_xlabel(\"Prediction Contribution\", fontsize=12)\n",
    "ax.set_ylabel(\"Error Contribution\", fontsize=12)\n",
    "\n",
    "fig.savefig(\"single_trn.png\", dpi=200, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "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>prediction_contribution</th>\n",
       "      <th>error_contribution</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>US Bond ETF_T-22</th>\n",
       "      <td>0.357729</td>\n",
       "      <td>-0.000426</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>INR_T-22</th>\n",
       "      <td>0.340033</td>\n",
       "      <td>-0.000584</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3M Libor_T-22</th>\n",
       "      <td>0.317645</td>\n",
       "      <td>-0.000776</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>EMEquity ETF_T-14</th>\n",
       "      <td>0.299084</td>\n",
       "      <td>-0.000438</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Crude Palm Oil_T-22</th>\n",
       "      <td>0.289143</td>\n",
       "      <td>-0.000404</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3M Libor_T-14</th>\n",
       "      <td>0.282654</td>\n",
       "      <td>-0.000586</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>White Sugar_T-14</th>\n",
       "      <td>0.256399</td>\n",
       "      <td>-0.000428</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Crude Palm Oil_T-14</th>\n",
       "      <td>0.245532</td>\n",
       "      <td>-0.000340</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Nikkei Ind_T-22</th>\n",
       "      <td>0.241579</td>\n",
       "      <td>-0.000817</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Corn_T-1</th>\n",
       "      <td>0.177843</td>\n",
       "      <td>-0.000232</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     prediction_contribution  error_contribution\n",
       "US Bond ETF_T-22                    0.357729           -0.000426\n",
       "INR_T-22                            0.340033           -0.000584\n",
       "3M Libor_T-22                       0.317645           -0.000776\n",
       "EMEquity ETF_T-14                   0.299084           -0.000438\n",
       "Crude Palm Oil_T-22                 0.289143           -0.000404\n",
       "3M Libor_T-14                       0.282654           -0.000586\n",
       "White Sugar_T-14                    0.256399           -0.000428\n",
       "Crude Palm Oil_T-14                 0.245532           -0.000340\n",
       "Nikkei Ind_T-22                     0.241579           -0.000817\n",
       "Corn_T-1                            0.177843           -0.000232"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.set_option(\"display.max_rows\", 10)\n",
    "\n",
    "contributions_trn.sort_values(\"prediction_contribution\", ascending=False).head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.4 Prediction vs. Error Contribution (Validation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "\n",
    "ax.scatter(contributions_val[\"prediction_contribution\"], contributions_val[\"error_contribution\"], s=80, color=\"orange\", edgecolors=\"black\")\n",
    "ax.hlines(y=0, xmin=0, xmax=contributions_val[\"prediction_contribution\"].max(), ls=\"--\", color=\"black\")\n",
    "ax.tick_params(axis=\"both\", which=\"major\", labelsize=12)\n",
    "\n",
    "ax.grid()\n",
    "ax.set_axisbelow(True)\n",
    "ax.set_title(\"Validation Dataset\", fontsize=12)\n",
    "ax.set_xlabel(\"Prediction Contribution\", fontsize=12)\n",
    "ax.set_ylabel(\"Error Contribution\", fontsize=12)\n",
    "\n",
    "fig.savefig(\"single_val.png\", dpi=200, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "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>prediction_contribution</th>\n",
       "      <th>error_contribution</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3M Libor_T-22</th>\n",
       "      <td>0.3106</td>\n",
       "      <td>-0.0277</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Nikkei Ind_T-22</th>\n",
       "      <td>0.2540</td>\n",
       "      <td>-0.0196</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Corn_T-22</th>\n",
       "      <td>0.1617</td>\n",
       "      <td>-0.0184</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Crude Palm Oil_T-22</th>\n",
       "      <td>0.2796</td>\n",
       "      <td>-0.0170</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>INR_T-22</th>\n",
       "      <td>0.3197</td>\n",
       "      <td>-0.0086</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Crude Palm Oil_T-14</th>\n",
       "      <td>0.2321</td>\n",
       "      <td>-0.0082</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Gold_T-22</th>\n",
       "      <td>0.1231</td>\n",
       "      <td>-0.0075</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3M Libor_T-14</th>\n",
       "      <td>0.2961</td>\n",
       "      <td>-0.0067</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Yuan_T-22</th>\n",
       "      <td>0.1133</td>\n",
       "      <td>-0.0062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>US Bond ETF_T-22</th>\n",
       "      <td>0.3782</td>\n",
       "      <td>-0.0061</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     prediction_contribution  error_contribution\n",
       "3M Libor_T-22                         0.3106             -0.0277\n",
       "Nikkei Ind_T-22                       0.2540             -0.0196\n",
       "Corn_T-22                             0.1617             -0.0184\n",
       "Crude Palm Oil_T-22                   0.2796             -0.0170\n",
       "INR_T-22                              0.3197             -0.0086\n",
       "Crude Palm Oil_T-14                   0.2321             -0.0082\n",
       "Gold_T-22                             0.1231             -0.0075\n",
       "3M Libor_T-14                         0.2961             -0.0067\n",
       "Yuan_T-22                             0.1133             -0.0062\n",
       "US Bond ETF_T-22                      0.3782             -0.0061"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.set_option(\"display.max_rows\", 10)\n",
    "\n",
    "contributions_val.sort_values(\"error_contribution\", ascending=True).round(4).head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "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>prediction_contribution</th>\n",
       "      <th>error_contribution</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Palladium_T-14</th>\n",
       "      <td>0.1320</td>\n",
       "      <td>0.0067</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Corn_T-1</th>\n",
       "      <td>0.1678</td>\n",
       "      <td>0.0044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>EMEquity ETF_T-14</th>\n",
       "      <td>0.3114</td>\n",
       "      <td>0.0041</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Nikkei Ind_T-1</th>\n",
       "      <td>0.1821</td>\n",
       "      <td>0.0027</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>US Bond ETF_T-1</th>\n",
       "      <td>0.0689</td>\n",
       "      <td>0.0025</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>INR_T-1</th>\n",
       "      <td>0.1539</td>\n",
       "      <td>0.0022</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Crude Palm Oil_T-1</th>\n",
       "      <td>0.0723</td>\n",
       "      <td>0.0021</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>EMEquity ETF_T-7</th>\n",
       "      <td>0.0741</td>\n",
       "      <td>0.0019</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Palladium_T-1</th>\n",
       "      <td>0.0931</td>\n",
       "      <td>0.0014</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>US Bond ETF_T-14</th>\n",
       "      <td>0.0813</td>\n",
       "      <td>0.0011</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    prediction_contribution  error_contribution\n",
       "Palladium_T-14                       0.1320              0.0067\n",
       "Corn_T-1                             0.1678              0.0044\n",
       "EMEquity ETF_T-14                    0.3114              0.0041\n",
       "Nikkei Ind_T-1                       0.1821              0.0027\n",
       "US Bond ETF_T-1                      0.0689              0.0025\n",
       "INR_T-1                              0.1539              0.0022\n",
       "Crude Palm Oil_T-1                   0.0723              0.0021\n",
       "EMEquity ETF_T-7                     0.0741              0.0019\n",
       "Palladium_T-1                        0.0931              0.0014\n",
       "US Bond ETF_T-14                     0.0813              0.0011"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.set_option(\"display.max_rows\", 10)\n",
    "\n",
    "contributions_val.sort_values(\"error_contribution\", ascending=False).round(4).head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 9. Recursive Feature Elimination (RFE)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.1. RFE - Error Contribution\n",
    "At each step, we remove the feature with the highest error contribution."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "rfe_error = pd.DataFrame(dtype=float)\n",
    "features_curr = features.copy()\n",
    "feature_drop = None\n",
    "\n",
    "for iteration in tqdm(range(len(features))):\n",
    "  \n",
    "  preds, shaps, shap_expected_value = get_preds_shaps(df=df, features=features_curr, target=target, ix_trn=ix_trn)\n",
    "  prediction_contribution, error_contribution = get_feature_contributions(\n",
    "    y_true=df.loc[ix_val, target], \n",
    "    y_pred=preds.loc[ix_val], \n",
    "    shap_values=shaps.loc[ix_val, :],\n",
    "    shap_expected_value=shap_expected_value\n",
    "  )\n",
    "\n",
    "  rfe_error.loc[iteration, \"feature_drop\"] = feature_drop\n",
    "  rfe_error.loc[iteration, \"n_features\"] = len(features_curr)\n",
    "  rfe_error.loc[iteration, \"contrib\"] = error_contribution.max()\n",
    "  rfe_error.loc[iteration, \"logloss_trn\"] = log_loss(df.loc[ix_trn, target], preds.loc[ix_trn])\n",
    "  rfe_error.loc[iteration, \"logloss_val\"] = log_loss(df.loc[ix_val, target], preds.loc[ix_val])\n",
    "  rfe_error.loc[iteration, \"logloss_tst\"] = log_loss(df.loc[ix_tst, target], preds.loc[ix_tst])\n",
    "  rfe_error.loc[iteration, \"ap_trn\"] = average_precision_score(df.loc[ix_trn, target], preds.loc[ix_trn])\n",
    "  rfe_error.loc[iteration, \"ap_val\"] = average_precision_score(df.loc[ix_val, target], preds.loc[ix_val])\n",
    "  rfe_error.loc[iteration, \"ap_tst\"] = average_precision_score(df.loc[ix_tst, target], preds.loc[ix_tst])\n",
    "  \n",
    "  feature_drop = error_contribution.idxmax()\n",
    "  features_curr.remove(feature_drop)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.2. RFE - Prediction Contribution\n",
    "At each step, we remove the feature with the smallest prediction contribution (i.e. the less important feature)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "rfe_prediction = pd.DataFrame(dtype=float)\n",
    "features_curr = features.copy()\n",
    "feature_drop = None\n",
    "\n",
    "for iteration in tqdm(range(len(features))):\n",
    "  \n",
    "  preds, shaps, shap_expected_value = get_preds_shaps(df=df, features=features_curr, target=target, ix_trn=ix_trn)\n",
    "  prediction_contribution, error_contribution = get_feature_contributions(\n",
    "    y_true=df.loc[ix_val, target], \n",
    "    y_pred=preds.loc[ix_val], \n",
    "    shap_values=shaps.loc[ix_val, :],\n",
    "    shap_expected_value=shap_expected_value\n",
    "  )\n",
    "\n",
    "  rfe_prediction.loc[iteration, \"feature_drop\"] = feature_drop\n",
    "  rfe_prediction.loc[iteration, \"n_features\"] = len(features_curr)\n",
    "  rfe_prediction.loc[iteration, \"contrib\"] = prediction_contribution.min()\n",
    "  rfe_prediction.loc[iteration, \"logloss_trn\"] = log_loss(df.loc[ix_trn, target], preds.loc[ix_trn])\n",
    "  rfe_prediction.loc[iteration, \"logloss_val\"] = log_loss(df.loc[ix_val, target], preds.loc[ix_val])\n",
    "  rfe_prediction.loc[iteration, \"logloss_tst\"] = log_loss(df.loc[ix_tst, target], preds.loc[ix_tst])\n",
    "  rfe_prediction.loc[iteration, \"ap_trn\"] = average_precision_score(df.loc[ix_trn, target], preds.loc[ix_trn])\n",
    "  rfe_prediction.loc[iteration, \"ap_val\"] = average_precision_score(df.loc[ix_val, target], preds.loc[ix_val])\n",
    "  rfe_prediction.loc[iteration, \"ap_tst\"] = average_precision_score(df.loc[ix_tst, target], preds.loc[ix_tst])\n",
    "  \n",
    "  feature_drop = prediction_contribution.idxmin()\n",
    "  features_curr.remove(feature_drop)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.3 Write Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "rfe_error.to_csv(\"gold_rfe_error.csv\", index=False)\n",
    "rfe_prediction.to_csv(\"gold_rfe_prediction.csv\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.4 Read Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "rfe_error = pd.read_csv(\"gold_rfe_error.csv\")\n",
    "rfe_prediction = pd.read_csv(\"gold_rfe_prediction.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.5 Plot RFE on Validation (Logloss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "max_iter_show = len(features) - 9"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "\n",
    "idxmin_prediction = rfe_prediction[\"logloss_val\"].idxmin()\n",
    "idxmin_error = rfe_error[\"logloss_val\"].idxmin()\n",
    "\n",
    "ax.plot(-rfe_prediction.head(max_iter_show)[\"n_features\"], rfe_prediction.head(max_iter_show)[\"logloss_val\"], lw=3, color=\"blue\", label=\"RFE - Prediction\\nContribution\")\n",
    "ax.plot(-rfe_error.head(max_iter_show)[\"n_features\"], rfe_error.head(max_iter_show)[\"logloss_val\"], lw=3, color=\"orange\", label=\"RFE - Error\\nContribution\")\n",
    "ax.scatter(-rfe_prediction.loc[idxmin_prediction,\"n_features\"], rfe_prediction.loc[idxmin_prediction,\"logloss_val\"], s=200, marker=\"o\", facecolors=\"none\", edgecolors=\"red\", lw=3, zorder=100)\n",
    "ax.scatter(-rfe_error.loc[idxmin_error,\"n_features\"], rfe_error.loc[idxmin_error,\"logloss_val\"], s=200, marker=\"o\", facecolors=\"none\", edgecolors=\"red\", lw=3, zorder=100)\n",
    "ax.set_title(\"Validation Set\", fontsize=12)\n",
    "ax.set_ylabel(\"Log Loss\", fontsize=12)\n",
    "ax.set_xlabel(\"N Features\", fontsize=12)\n",
    "ax.set_xticklabels([-int(x) for x in ax.get_xticks()])\n",
    "ax.grid()\n",
    "ax.legend(loc='center left', bbox_to_anchor=(1, 0.5), fontsize=12)\n",
    "ax.tick_params(axis=\"both\", which=\"major\", labelsize=12)\n",
    "\n",
    "fig.savefig(\"rfe_logloss_val_.png\", dpi=200, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "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>feature_drop</th>\n",
       "      <th>n_features</th>\n",
       "      <th>contrib</th>\n",
       "      <th>logloss_trn</th>\n",
       "      <th>...</th>\n",
       "      <th>logloss_tst</th>\n",
       "      <th>ap_trn</th>\n",
       "      <th>ap_val</th>\n",
       "      <th>ap_tst</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>104</th>\n",
       "      <td>Nifty_T-7</td>\n",
       "      <td>16.0</td>\n",
       "      <td>-0.00313</td>\n",
       "      <td>0.003295</td>\n",
       "      <td>...</td>\n",
       "      <td>0.274596</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.74032</td>\n",
       "      <td>0.700379</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 9 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    feature_drop  n_features  contrib  logloss_trn  ...  logloss_tst  ap_trn  \\\n",
       "104    Nifty_T-7        16.0 -0.00313     0.003295  ...     0.274596     1.0   \n",
       "\n",
       "      ap_val    ap_tst  \n",
       "104  0.74032  0.700379  \n",
       "\n",
       "[1 rows x 9 columns]"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rfe_error.sort_values(\"logloss_val\", ascending=True).head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "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>feature_drop</th>\n",
       "      <th>n_features</th>\n",
       "      <th>contrib</th>\n",
       "      <th>logloss_trn</th>\n",
       "      <th>...</th>\n",
       "      <th>logloss_tst</th>\n",
       "      <th>ap_trn</th>\n",
       "      <th>ap_val</th>\n",
       "      <th>ap_tst</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>104</th>\n",
       "      <td>S&amp;P 500_T-14</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0.304652</td>\n",
       "      <td>0.003227</td>\n",
       "      <td>...</td>\n",
       "      <td>0.332467</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.662959</td>\n",
       "      <td>0.63836</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 9 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     feature_drop  n_features   contrib  logloss_trn  ...  logloss_tst  \\\n",
       "104  S&P 500_T-14        16.0  0.304652     0.003227  ...     0.332467   \n",
       "\n",
       "     ap_trn    ap_val   ap_tst  \n",
       "104     1.0  0.662959  0.63836  \n",
       "\n",
       "[1 rows x 9 columns]"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rfe_prediction.sort_values(\"logloss_val\", ascending=True).head(1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.6 Plot RFE on Test (Logloss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "\n",
    "idxmin_prediction = rfe_prediction[\"logloss_val\"].idxmin()\n",
    "idxmin_error = rfe_error[\"logloss_val\"].idxmin()\n",
    "\n",
    "ax.plot(-rfe_prediction.head(max_iter_show)[\"n_features\"], rfe_prediction.head(max_iter_show)[\"logloss_tst\"], lw=3, color=\"blue\", label=\"RFE - Prediction\\nContribution\")\n",
    "ax.plot(-rfe_error.head(max_iter_show)[\"n_features\"], rfe_error.head(max_iter_show)[\"logloss_tst\"], lw=3, color=\"orange\", label=\"RFE - Error\\nContribution\")\n",
    "ax.scatter(-rfe_prediction.loc[idxmin_prediction,\"n_features\"], rfe_prediction.loc[idxmin_prediction,\"logloss_tst\"], s=200, marker=\"o\", facecolors=\"none\", edgecolors=\"red\", lw=3, zorder=100)\n",
    "ax.scatter(-rfe_error.loc[idxmin_error,\"n_features\"], rfe_error.loc[idxmin_error,\"logloss_tst\"], s=200, marker=\"o\", facecolors=\"none\", edgecolors=\"red\", lw=3, zorder=100)\n",
    "ax.set_title(\"Test Set\", fontsize=12)\n",
    "ax.set_ylabel(\"Log Loss\", fontsize=12)\n",
    "ax.set_xlabel(\"N Features\", fontsize=12)\n",
    "ax.set_xticklabels([-int(x) for x in ax.get_xticks()])\n",
    "ax.grid()\n",
    "ax.legend(loc='center left', bbox_to_anchor=(1, 0.5), fontsize=12)\n",
    "ax.tick_params(axis=\"both\", which=\"major\", labelsize=12)\n",
    "\n",
    "fig.savefig(\"rfe_logloss_tst_.png\", dpi=200, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.7 Plot RFE on Validation (Average Precision)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "\n",
    "idxmax_prediction = rfe_prediction[\"ap_val\"].idxmax()\n",
    "idxmax_error = rfe_error[\"ap_val\"].idxmax()\n",
    "\n",
    "ax.plot(-rfe_error.head(max_iter_show)[\"n_features\"], rfe_error.head(max_iter_show)[\"ap_val\"], lw=3, color=\"orange\", label=\"RFE - Error\\nContribution\")\n",
    "ax.plot(-rfe_prediction.head(max_iter_show)[\"n_features\"], rfe_prediction.head(max_iter_show)[\"ap_val\"], lw=3, color=\"blue\", label=\"RFE - Prediction\\nContribution\")\n",
    "ax.scatter(-rfe_prediction.loc[idxmax_prediction,\"n_features\"], rfe_prediction.loc[idxmax_prediction,\"ap_val\"], s=200, marker=\"o\", facecolors=\"none\", edgecolors=\"red\", lw=3, zorder=100)\n",
    "ax.scatter(-rfe_error.loc[idxmax_error,\"n_features\"], rfe_error.loc[idxmax_error,\"ap_val\"], s=200, marker=\"o\", facecolors=\"none\", edgecolors=\"red\", lw=3, zorder=100)\n",
    "ax.set_title(\"Validation Set\", fontsize=12)\n",
    "ax.set_ylabel(\"Average Precision\", fontsize=12)\n",
    "ax.set_xlabel(\"N Features\", fontsize=12)\n",
    "ax.set_xticklabels([-int(x) for x in ax.get_xticks()])\n",
    "ax.grid()\n",
    "ax.legend(loc='center left', bbox_to_anchor=(1, 0.5), fontsize=12)\n",
    "ax.tick_params(axis=\"both\", which=\"major\", labelsize=12)\n",
    "\n",
    "fig.savefig(\"rfe_ap_val_.png\", dpi=200, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "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>feature_drop</th>\n",
       "      <th>n_features</th>\n",
       "      <th>contrib</th>\n",
       "      <th>logloss_trn</th>\n",
       "      <th>...</th>\n",
       "      <th>logloss_tst</th>\n",
       "      <th>ap_trn</th>\n",
       "      <th>ap_val</th>\n",
       "      <th>ap_tst</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>101</th>\n",
       "      <td>INR_T-14</td>\n",
       "      <td>19.0</td>\n",
       "      <td>-0.000959</td>\n",
       "      <td>0.002888</td>\n",
       "      <td>...</td>\n",
       "      <td>0.256222</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.747679</td>\n",
       "      <td>0.72811</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 9 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    feature_drop  n_features   contrib  logloss_trn  ...  logloss_tst  ap_trn  \\\n",
       "101     INR_T-14        19.0 -0.000959     0.002888  ...     0.256222     1.0   \n",
       "\n",
       "       ap_val   ap_tst  \n",
       "101  0.747679  0.72811  \n",
       "\n",
       "[1 rows x 9 columns]"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rfe_error.sort_values(\"ap_val\", ascending=False).head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "scrolled": true
   },
   "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>feature_drop</th>\n",
       "      <th>n_features</th>\n",
       "      <th>contrib</th>\n",
       "      <th>logloss_trn</th>\n",
       "      <th>...</th>\n",
       "      <th>logloss_tst</th>\n",
       "      <th>ap_trn</th>\n",
       "      <th>ap_val</th>\n",
       "      <th>ap_tst</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>106</th>\n",
       "      <td>INR_T-14</td>\n",
       "      <td>14.0</td>\n",
       "      <td>0.366041</td>\n",
       "      <td>0.003611</td>\n",
       "      <td>...</td>\n",
       "      <td>0.322916</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.664506</td>\n",
       "      <td>0.655899</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 9 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    feature_drop  n_features   contrib  logloss_trn  ...  logloss_tst  ap_trn  \\\n",
       "106     INR_T-14        14.0  0.366041     0.003611  ...     0.322916     1.0   \n",
       "\n",
       "       ap_val    ap_tst  \n",
       "106  0.664506  0.655899  \n",
       "\n",
       "[1 rows x 9 columns]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rfe_prediction.sort_values(\"ap_val\", ascending=False).head(1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.8 Plot RFE on Test (Average Precision)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAh8AAAEcCAYAAABnIdbiAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAB/d0lEQVR4nO2dd3gVVfrHP28K6SGEhEBCCV2QDnZFWAt2RFzxZ9t1VVZWXde61rXvuuqqay/r2rAhthV7AXulN6WGFkogJIQkpJ7fH+fezNyW3IQkpLyf57lPZs6cmXsmbb73rWKMQVEURVEUpbmI2NcLUBRFURSlfaHiQ1EURVGUZkXFh6IoiqIozYqKD0VRFEVRmhUVH4qiKIqiNCsqPhRFURRFaVZUfCiKoiiK0qyo+FCUOhCR3a5XtYiUuvbPbsD15ojIhXXMuUBEfhGRIhHZKiLvi0hSGNceJyIb67smRVGU5iRqXy9AUVo6xphE77aI5AAXGmM+bar3E5Ejgb8Dxxlj5otIKnByU72foihKc6OWD0VpICISISLXichqEdkhIjM8QgERiRWR6Z7xAhH5SUQyROQu4AjgEY/l5JEglz4A+M4YMx/AGJNvjHneGFPkuXaMiNwnIus9VpEnRCRORBKAD4BMl2Ums3m+G4qiKOGj4kNRGs5lwKnAkUAmsBN41HPsd0BHoAfQGbgYKDXG3Ah8BVxqjEk0xlwa5Lo/ABNE5DYROUxEYvyO3w0MAEYA/YAs4G/GmGLgeCDXc+1EY0xuo92toihKI6HiQ1EazsXAjcaYjcaYMuBW4HQRiQIqsKKjnzGmyhgz1xizK5yLGmO+Ak4DRgHvATtE5H4RiRQRAaYCV3gsIkVYF82ZjX53iqIoTYTGfChKw+kFvCUi1a6xKiADeBFr9XhVRFKA6VihUhHOhY0xHwAfiEgEMB54HfgVeAuIB+ZaHQKAAJF7fTeKoijNhFo+FKXhbACON8akuF6xxphNxpgKY8xtxpjBwKHAScB5nvPCbiVtjKk2xnwGfA4MAbYDpcD+rvfs6AqK1TbViqK0eFR8KErDeQK4S0R6AYhIuohM9GyPF5GhIhIJ7MK6YbwWkq1An1AXFZGJInKmiHQSy4HYuJLvjTHVwNPAAyLSxTM/S0QmuK7dWUQ6Nv7tKoqiNA4qPhSl4fwb+B/wsYgUAd8DB3mOdQVmYoXHcuALrCvGe97pIrJTRB4Kct2dwEXASs/504F7jTEveY7/FVgFfC8iu4BPgYEAxphfgFeANZ4sG812URSlxSHGqJVWURRFUZTmQy0fiqIoiqI0K80mPkQkVUTeEpFiEVknImeFmBfjKZq0VUTyReRdEcmq73UURVEURWmZNKfl41GgHJuGeDbwuIjsH2Te5cAhwDCcwk0PN+A6iqIoiqK0QJol5sNT9nknMMQYs8Iz9iKwyRhznd/cx4EiY8y1nv0TgfuNMQPrcx1FURRFUVomzVVkbABQ6RUMHhZi0wf9eQb4tydKvwBr3figAdfxIS0tzWRnZ9d/5UBxcTEJCQkNOrel05bvDdr2/em9tV5a0/3NnTt3uzEmfV+vQ2lbNJf4SMSmDLopBIK1CF+JLd60CVstcjHg7X9Rn+sgIlOxpajJyMjgvvvua8ja2b17N4mJiXVPbIW05XuDtn1/em+tl9Z0f+PHj1+3r9egtD2aS3zsBpL9xpKBoiBzHwVisH0xioFrsZaPg+p5HYwxTwFPAYwZM8aMGzeuQYufM2cODT23pdOW7w3a9v3pvbVe2vr9KUpdNFfA6QogSkT6u8aGA0uDzB0BPOdpmlWGDTY9UETS6nkdRVEURVFaIM0iPjytvt8EbheRBBE5DJiIU/HRzU/AeSLSUUSigT9hW4Rvr+d1FEVRFEVpgTRnqu2fgDhgG7b88zRjzFIROUJEdrvmXQ3swcZ+5AEnAJPquk4zrF9RFEVRlEaguWI+MMbkA6cGGf8KG0jq3d+BzXCp13UURVEURWkdaHl1RVEURVGaFRUfiqK0LoyBwuVQXbWvV6IoSgNR8aEoSuvixz/Ce4Ph86OsEFEUpdWh4kNRlNZFzkv267YvoGTjvl2LoigNQsWHoiith8piqCpx9ks37bu1KIrSYFR8KIrSetizzXdfLR+K0ipR8aEoSutBxYeitAlUfCiK0npQ8aEobQIVH4qitB7KVHwoSltAxYeiKK0Hf8tHqYoPRWmNqPhQFKX1sGer775aPhSlVaLiQ1GU1kNAzMcmMNX7Zi2KojQYFR+KorQe/MWHqQwcUxSlxaPiQ1GU1oN/wCkEul5MNeR+CPnzm2dNiqLUGxUfiqK0HoJZOfzFx8onYM7x8NEBKkAUpYWi4kNRlNaBqYayvMBxf/Gx5VPP/CpY+2LTr0tRlHqj4kNRlNZBWX7w4FL/dNvSXGd784dNuyZFURqEig9FUVoHweI9INDy4RYfu5ZD8bqmW5OiKA1CxYeiKK0Dn3gPcTbd4sNUQ+lm3/M2f9Sky1IUpf6o+FAUpXXgFh/J+znbbvGxJ8+m37rJ/aBp16UoSr1R8aEoSuvALT5SRznbJRvBGLvtdrl42fIZVJU37doURakXKj4URWkduGM+EvtCdLLdri6Dsh12u3RT4HmVRbD9u6Zfn6IoYaPiQ1GU1oHb8hHbBeK7O/vejJdglg+AzbW4XoyB6qq9X5+iKGGj4kNRWjvtxaXgLz7iXOLDG/dR4hIfKUOd7dz3g6fpFiyGt7rBrIGwc0GjLldRlNCo+FCU1szaF2FmCnw+oe03WHN3tI3N8LV8eMWH2+2SfQ5ERNvtgsXwyeGw42ffay680V5392qYPQF2rWiatSuK4oOKD0VpzSz9B1SVwpaPIX/uvl5N0+K2fMR0CSE+XJaP5IFWgHjZ/h18dCD88oDnnFzIfc/3+p8fA8UbGn/tiqL4oOJDUVopEaYcilyf1P2LbbU1ymqJ+QgmPuKyYMyjMPg6iOjgGTQw/1rYtRLWPh9oLSpZD3NO0BgQRWliVHwoSislvnKd7V/iJVSwZUtg7Ysw/69QtKph51ftgYpddluioENKcPFR4nK7xGVCVByM+AecsAQ6jbTjphIWXgern3Hm9r3QcdEULoEdPzZsnYqihIWKD0VppSRUrPEd8K/s2VLYuRC+Ow+W3wPvDYEld9U/SHaPq6FcbDpIBMT3cMZ2r7HX9DaekwhrHfGS3B8OeMLZ3/CmjfMAiE6B0Q9B1snO8eK1wdexOwd2hzimKErYqPhQlFZKQoXfQ7ClWj52/ORsV5fBopvgozFQujX0OV7y50Hed74ulxiPqEjq71gritdC4VJnTmxXiIjyvVbagdBzSuB7ZJ9tLSTxvZyx4vWB87Z8CrMGwLv9YduXda9dUZSQqPhQlFZKYmUrsXwEa+xWsNjGXNTGti/hw9HwyaGw5E5n3GvRiIyBlOHO+Kb/OdtxmcGvOeLvjmDx0u9C+zXBLT781lxdAT9far+aKtjwVu1rVxSlVlR8KEorpdVYPtwP8sS+znbhstrP2/K5s73xbWfb7U7pfEDwOaHER2If6H+Js586GjqNsNsJPYOvGWD1f2DXr67jOaHXrShKnaj4UJTWSPlOYqq3+461VMtHietB3mOys11UR00Nt6vFTYxLfKS6xIe7SFh8VujrDv0bdDnSCpTRDznjbstHicvtUrELFt3iew0VH4qyV0TVPUVRlBZHweLAsbI8G3QZ2SHw2L7EbUXodowNPIW6xceeEOIjlOXDTSjLB0CHTnD0nMBxf7eLMSACy/7pBLJ60aBTRdkrms3yISKpIvKWiBSLyDoROSvEvA9EZLfrVS4ii13Hc0Sk1HX84+a6B0VpMQQTH+BbBbQlUF3pW38k7VCIjLPbZTuchnDBCCk+Mpzt5EEQlRA4pzbxEYoOqRAZb7crd0NFgV3DL/cHzq0ohPKC+r+HoihA87pdHgXKgQzgbOBxEdnff5Ix5nhjTKL3BXwLvO437WTXnGObfOWK0tIIJT5aWtxH6SanFklsBkTF2ywVL0UrQ5/rtjZ0SHW2E/s42xGR0GlU4LlxtbhdQiESaP3Y9oWtMQKQMgySBriO59T/PRRFAZpJfIhIAjAZuNkYs9sY8zXwP+DcOs7LBo4AXmjyRSpKa8ItPryf1qHlxX24XS7eB7v7AV5bLxW35WPs25B5Igy8Aroc4TsvmOslvgGWD/cawa595yJnv9txvgGzu3Ma9h6KojRbzMcAoNIY4/5PsxA4so7zzgO+Msbk+I2/JCIRwHzgGmPMwmAni8hUYCpARkYGc+bMacDSYffu3Q0+t6XTlu8Nmu/+UsrmkVT+K7kJp1AVEcQN0ECSy5eSuucHNiecTFlkuh00hsN3LKz5490RNZTOVT8AsGLRHHJXpzTa++8tGSUfM8izva04jmVz5tB7VwzeR/y6JZ+wdn3PgPOKiwqgPB8Ag/DF0nKQq6EI+MK3xkaX0gQG+53/9bw1VEbU4tIJwYDCKLyyZeWCz+hUNpc0z/6yzVF0LI/Ga1NZtfATNjbwe93W/+4UpS6aS3wkArv8xgqBpDrOOw+402/sbGAeIMDlwEcisp8xpsD/ZGPMU8BTAGPGjDHjxo2r98IB5syZQ0PPbem05XuDZrq/HT/Dx9eCqaJvxzwY+2bjXLd8J7wzCSoKyDZfw0nLbW2L4vXwTrGdE51C5/7HwhIrPgZ0T2LA8HGN8/6NwZKvocBudul9AF1GjoM16+D7lwDolVpGr8PHBZz27WdvWKEBSEwa48YfFfo9inrAu3c4+xExHD7+FOtGqS9Lv4WFswDon9kBNmyCMnto8KFnQu4HsMDWE+nXNYp+YwLXHg5t/e9OUeqiuWI+dgPJfmPJ1Px7CUREDge6AjPd48aYb4wxpcaYEmPMP7D/2o4IcglFaR4W3+rENWx8u/HM8Vs+tUGPYCt4rnjEbrtdLilDfeMbmirmY0+ezf6oL263i7eCaPJAZyyE2yW6eqez485uCUZiH5vB4iUus2HCw71GgIJFzvojoiF5P0jIdo5rzIeiNJjmEh8rgCgRcUWaMRxYGmI+wO+AN40xu+u4tsFaQRSl+dnxs29bdoxvw7K9IfdD3/0ld0JZPuS+74ylDIW4bs5+U4iPhTfBm11gzon1FyB1xXwUrQzsLAt0qCpwduoSHyKQOsbZb2i8B/gWGtv2hbOdPNgKkMTezpi3/0vRKvuzyZ/f8PdVlHZGs4gPY0wx8CZwu4gkiMhhwETgxWDzRSQOOAN4zm+8p4gcJiIdRCRWRK4B0oBvmvQGFCUUS24PHFvzjE0xrS9Ve5zzjIHNH/keryiAT8fCysecsbRDfNNKGzvg1BjH4rL5g/p/2vepbpptv8akQkxnu11V4tuJ1oOP5SOmDvEBvkGnDUmz9eIOOK12Nb9LGeY5nu2M7c6x358vJ8Kim+GjA2HV0w1/b0VpRzRnqu2fgDhgG/AKMM0Ys1REjhARf+vGqVh3ymy/8STgcWAnsAk4DjjeGFP/yDJF2Vvy58Gmd5396BT7tXSznzUkDHI/gJmpMGugrYtRuNSmqYLt0OrF1TwtP2YM9JrStJaPsu22poWXkg3hn2uqg1s+wM/6Eeh66VBd4OzUZfkA6D6JGgNoRi3xIXURlwkSGTjeySM+YtJctUCKIO8bp0y8qYQfp8LPlzdMfCpKO6LZxIcxJt8Yc6oxJsEY09MY87Jn/CtPPQ/33FeMMb2M8bXxGmOWGmOGea7R2RhzlDHm5+a6B0WpobIU5l/j7Pc4HfpPc/ZXPVWPa5XADxdBValtDb/oZtjscrl0Pw06H+R7TrfjWJJ6l3UFxGZQ8+Aty7PNzxqLolW++8X1EB97ttkutmCFWbQr7KsO8RFdX/HReQwc+y0c+S70vTD8NfoTERW8Roi3gZ2Ir+tldRBLx4qHYP7VDV+DorQDtLeLotSX3TnwyWGw1dX4bOjfnO6oYC0ZwdqyB+OX+x0rB8DaF2D1f539zONh1P0gnuS0zBNh7NtUi6eMekSU7wO6dEu9bqdWdvuJj/pYPkJZPQCSa6/10aE+Aade0g6GrJNs4bG9wX+t4LhdwNf1sn5G8PGVj8Mev947iqLUoOJDUepD/nz4aAzsdAUXDrrGBn4m9oGux3gGDax5Lvg11s+EH/8IW7+wQmHZ3b7HTTXsWu7sdzsW0g+FCd/DEW/ZgluRMb7nNFXch7/lw10qvS5qEx91WT7cAafhxHw0Jv5rje0Cca6S7m6R4a1+CvCbT51Gd9XlNvZHUZSgqPhQWg+/PABvZsDiO+qe21TMv8bpRxIRDQc8DiP+6Rzv8wdne8MbgecXLIVvpli3zGfj4MMxUOmp2REsULLjEIjvbrdTR0OPU62lw5+mivsIEB+NZPmoo8ppgywfjUWCX9Ezr8vFizdw1mesDyT1hQGXOmMrH4fqqkZfnqK0BVR8KK0DU21TPvdsg8V/a7oy4qv+A58fA7kfBR6rroDtrsSq33wK/S/2rSmRdSJEeKwSBYsCH96/PuCbWup2txz0DGSe4Ds/87jw1u0WLnsa8XvTaG6XbN9jSf1c89babrwu6h3z0Zj4CyW3ywUgoTcBeC1evc5wMnmK19kMITfGwOzZ7Pf3v8OBB8KQIXDYYXDFFbB8OYrSXlDxobQOSrfYtEwv7loXjcXq/8KPF9niXj/9MfB4wWLHzJ7QC7qMDZwTnWTdJF42uKqdlm6FtdODv3fXY6DbBBjml7rbbUJ4a3eLj5KmtHw0ktslKh7iPRYGU2UDbV3UO9ulMYmvQ3wEs3x4xUdkLPS5wBlf8aizPXMmDB4Mv/kNXT/5BH76CZYuhW+/hQcftMeOOgoWLGiEm1CUlo2KD6V14H6QAWya1bjXz/sGfrrY9/3KC33n7PjB2fbPPnHTY7Kz7RYfKx93sj86HwjHL4Dsc+z8Q573FMsa7bhukgdBehCBEwy326WxLB9l+TX9VZyxPN84h9pw1wQJFsSZvJ+zvXOBs11ZTKTxvEdEDETV1YWhkfF3u3Tyt3xk++5LBHT9jbPf/2Jqso82f2gF3F13wW9/C7/8Uvt7f/65tYR8FMTypihtCBUfSuugxC9zZMsn4T8EvVRXwo8Xw5yTfTNRitfDV6cFpqjuXu27vz1M8ZF1spOZsuMHay2oLPUtDrbfVdBpOBz6Ihwx01c8HPgUHDcPjpsLkR3Cu7eGWj6Mge0/wvo3Au/f3+pRc/0wrB/G1G75ACvAvOz43tl2d7ON7dLwUukNJaGXU+sjooMVgW46pEKUqzpA6hjf8u6JvX3dZ3f/CW66ydlPSmTTqRNh9mxr5XjnHTjtNIjw/DsuKYFJk6xlRFHaKGGJDxHpLSIvi8gyEVnvfjX1AhUFCExbrSy22SL1YeM7sOpJyJ0FS//hjM+93PeB58VffLgfkGm1iI+YVMgY7+xveAtyplurAVh3Q4/TQp8fEQmpIyEqLvQcf+pr+agssUGvH4yEjw+Cr0+Hb870neMf7+ElHPGR+54twgUQlWCLc/mTdrCz7RZ2/uKjuYmKh/1vsrVJht0emFnkX+uj67EE0N9jRdsF3PuJM35Ad7hvN8lnLYNDR8Lw4XDKKfDGG7BwIfTyiLTSUvjjHxvWT0dRWgHhWj5eBqqBq4Bz/V6K0vT4u13Aioj6ULDQ2d45z341Bra6Cum6LRruOITynbDrV7stUdBpVO3v5RYXS+6An/7k7A+8PHjGyt7gtnzszrHiIhjGwLoZMGs/m+7r/p5seNO3n0xIy0cdQacVRfCTq+Bar/8Lbr1wWz52zoMqj0vKLT6aO83Wy7Bb4fR8GPzX4MfdvWS6nxJ4vNsEiEmHLwCvQWlgX5i2ERIhqWIlfHu2b/DxkCHW3RLjETvz58MPP/hfWVHaBOGKj/2B84wxHxhjvnC/mnJxilKDv9sFbGnz+nwyLFrpbBcut+eW5jrlw6M7QvZZrvkuy8cOlwk8ZVjdVonup+JTddR4ym3HpEPfC0Kd1XBiM5wsi4oC+PnSwDl78uDzo2yqbygBMe9Kx/3iFh/e0vFQt/hYeINjHYlJhxF3B58Xmw6Jfe12dbkT91G2jy0fXmpz9wy7EwZeAQc/79tXxktENPScAp+5xs7sDW4jSu57sOgW3/MGDoSzz3b2H3+8IStXlBZPuOLjS2BkUy5EUWolWLXQ4nU+vU7qxC0+Kotsmqv7/I6DnYch+Lpd3G4Bt7sgFHFdIf0w37EuY+Goz6BDx/DXHC4RUTDc9ZBf86xvlVSwAbVuK09Muq2cetIvTlDnruWw8gm77Xa7ZBzpbNfmdsn7zjfDY/SDjigKhvt76Q3o3ddul3CIz4TR90Of80LPiZ0AHk8bMUC/bwPnLL3TNygZ4A+uWjFffbW3K1WUFkm44iMH+FBEnhKR292vJlyboji43S5dXA/CcLNejAl0IxQuc5qCAXTcP7T4CDfTxc2Ie2xGR8ZRtibIUXNsJdSmou8F0Nv1MPz5EtjpcauU7YCN/3OODbwcTl4B+10ByQNhiCsgcvEtNtPF/f3qMs7Zrq2/y8LrAY81qtvx1uVSG+7v5XZPTE1rEB/hYFxZM2lAhMcV1nGIbQro5YcLfYOEBw50tne6iq0pShsiXPGRAMwCooEerlf3JlqXojhU7LKuBLCpl33Od465u8rWhn93VvCID7flY39PDQePub1kgy1+ZYyv+Kgt2NRN+iFw0nI46lPoelTTZ22IwAGP2aqoYLOBfr7Mbm94w3H9dD7YWiQ6pDjnDrzcVukEG9/y41QnQDYixtdCEcrtUl4AeV97F2PXUtc912X52FcxH41BnMs1507M2u9KlnW62UnZLd9p42+8LsTdribfsbFNvUpF2SeEJT6MMeeHeP2h7rOVds3WL2DNC04wYUNwf9KO72Ebq3nbzG//zhbvqgu3y8VLgOVjsC0SFe/pauptCV+81ooXsLEPSf0bdBvNQlSCTd31pvrmfWUtCjmvOHOyg1gjImNg5L+cfXdp+KS+vqmypSHcLltn24JhYAMygxXj8idluFMRdvcaKzx8Yj7S675GSyUzE6K8KdfAemxsTvZZVEYkw8HPOnNzZ9mGggDvugR1L9f3XVHaEGHX+RCR/iLyNxF50vO1Bf8HVloEOxfCZ+Ph+9/Bsnsafh3/ehGxaZDmjacw4WW9BBMfu4JYPiDQ9bLdlWLb+UBH+LRUkgdCtitocf41sM0TGy4R0PO3wc/rcWpwN0liP2uB8Aqash3Bs2k2f+xsdwuSfhqMyA6Q6soc2v5D23G7JCTAZFfBuY+B/pc4qbsZ42DAZc7xuZdD0Tp41BUz87vfNcdKFaXZCbfOx8nAXGA/IB8YCPwsIkFyzBTFQ+4H1Pj/c99r+HXcmS7e6pPdJzpjG9+p+xrB0kZ3/OSb6eJNV/UXH9tcQX/hulz2NYOudrbzvqbm59BlnG9NEH8OeBTisnzHkvrb2iPxrvFgQadbGiA+wLqBvOz4oe24XQD+5EqxngMsz/Y9PuIfzu9beSGcOwJ+9aR0JybCOec0/RoVZR8Q7ke4vwMTjTFnGWOuN8acDUz0jCtKcLy1NMA2WXN3+KxPt093pkt8EPGx5ROnM6yX3Wth/l/pWLbI7gezfFS7mpl1HOzEJyS5xEfRal/hlHFU+Ovel6QMsQGf/tQVANqhk687AJwmcPE9nDF/8VG02qmLEpXoKyjqwi3o8r52Yk2gdbtdAI44wr7A6r//+z1cdRWxW7bYsagEOOhZWB0BDwDvFDjnXn45JDVzaXlFaSbCFR/dAf+cr6/RgFOlNvJd4qOqFIo8n+hWPwszEmDWYNvvxF84+BOsTHdSPysYwAZWbnZVkSzLh48PheX3MDT/Bk/mhkt8RAQpWe51uQAk9HG2N73rBFhGd4T0Q2tfa0ti8LW++xHR0HNy8Lluuh0D+11ptyUSMjx9S+Jdf+7+Qaduq0fG+PDLwoNv0Om2L5y4kehkG4PTmhGxDeX6eQRcdTXcfz8HnXWW7Wp73HFw5EXwt2prW/ZyQAe4PkitFkVpI4QrPhZgq5u6udIzriiBlBcElifPn2+/LrndNljbtdxW/ny7h29ApD/B3C7gKeTlYePbzva8K2CP/WQZZYptB1yfmhWuJmBe3OLDbflwn9dtgn2Atxa6HOlbibPbcb49SGpj5H0w7kPb/C55gB3zsXz4iQ93vEewcuO1Ed8zsHMstH6Xi5cuXeDrr+Egx8IjxtjeLR995LhZvIwDLi2HDa826zIVpTkJV3xMAy4UkVwR+UFEcoGpnnFFCWTn/OBju3N8u52CTTWce1noaqVut4s76yLL5XrJnWUbx216z8ka8LL6GZuuC7aYlrtOiBevFQV8Yz7cuJuFtQZEbCVOb+rwwD/X79zMCdZ94yWU26W6ArZ+7uzXJ97D+16/+QT6XuQ0dIPwsmVaCxkZ8M03tonchAmBx2Nj4fzzYcblcBEQRf0K6ClKKyOsBhPGmF9EZBBwCNANyAV+MMZU1H6m0m5xu1y87JwH2+Y4+4n9rFWjutxmUJTnO9UwK0s9/U/EViL14jb9dx5jgydLN9vzvzvPyepw437PpP6+Vg4v7rGYVJtS660tAnYdmUFiKFo6mRPgRE86ccf9ap9bF27xkfcVLPVUVN2z1RF3Cb0alooc2wUOegoGXQVL7mLXpp9JHtrGahhGRtomcqecwnevv84h6em2g21ysu3rkpICuR/BnH/b+buW79PlKkpTEnZ3K2NMJYFxH4oSnGDiI3++7wOs7wW226v3E15xjhUfBUvg0yMBAwc948QAxGb4xgBIBGSdYjvVAqxzuW5i0m1Ko39gpDtWxIs706VmXl/IdznhOx/QetM+91Z0eHELv8JlnmqmfnQ9du+KqSUPhENfYN6cOYxLP6Th12nhlKWnw7hxgQc6DnK2vf2Hmro4naLsA0KKDxFZbowZ5NneQE2uni/GuGsIK4qHnXMDxyoKbHt5LxnjbHZDjfhYB6mjYe2L1goC8P3vnflul4uX/hfDmv86zdC8HPAo5H0Dv/7bdzypv60sGRlrA1XBN9PFS2IfX/GReWLw+2xPdBwEMWlOwbVg9Dqj+dbTFonvYTNgKovt30BZXosVvXPnzu0QFRX1NHA4EFnXfKXdUC0iWyorK28bNWrUR6Em1Wb5uMi1rcnmSvhUFMGuFXZbIiD1AKd0dmWR/RqVaIWGW1DszrFf3ZkpXnM+OGm2bjqNgFPWwo4foWSTdQF0HmNTcTt0Di4+IiJtzxVvF9Vgbhj/uI/WFu/RFEQlwFGfw/rXfdOUvXQ+GLoe3fzrakuI2N9Nr/AtXN5ixUdERMS05OTkw3r16lUQERFRj/bSSlumurpaSktLO+bk5Dwyb968S0MJkJDiwxjztWs7iCNdaRSqq2xMQkIvp55CU7LpfVjzjG1A5q6V0ZjsXEiNoSx5kO3u6u6NApB+uM0c8fa3ACeldneQgmDgm+niJj4L4icFjnc5gkpJsBkvXrzf4+TB4YuP2AzfKpztmZShTdscT7F/M17xsWu5b0fhFkRkZOT5mZmZxSo8FDcREREmISGhNDs7m9WrV98CBBUf4VY4vVJERni2DxaR9SKyVkTarlO2ufjlfvj8aHhvcO2tyhuD6kr4/jzbwvvLU2HV0w27zqb3YM3zga4OL+7iYqmjodPIwDkZ4+xXt+WjOCd491kvwdwutRERzY5Yv4qk3mDI/tPsJ/m4TMg+K/DcLkc4ZdSzz275JdWVtoN/3EcLxRjTsUOHDpp0oAQlLi5ujzGma6jj4QacXgE849n+B3A/UAQ8CLSSetMtlDWeb2t1hS3jHazpV2PhzQrx8uNUunW8EltYIExyP4AvTrLb5QWw3+WBc9zBpp1GBRcfXcbbr/6Wj9LNtiAZ2GJgbvN+MLdLHeyIPYyMUk8aaHSyDUQF6HI4nLbVNjWLCPJnkDwQfvO5LYyWrV5HpRlJdomPlp3xIqLBsEoIPBaxkJ/awv0419EYUygiScBw4GFjzDPYHi9KQyneALtcBYZ8UjubAHexLg8DC++HnHoUM1rhanoVql+LO1AzdZR9kLuzVKISHTeGv+XDXZgsZTgM8FR5jEqwrpp6siPmIOs2gcBMjKiE4MLDS8aR0G8qRMXX+30VpcG0HvGhKA0mXMvHBhE5FNgf+NIYUyUiyUA9GnQoAWz5xHe/vKBp3684UHwAsPhv0GtK3Sl9pZth8wfOfv7cwFTAylLXP0yxAaERUbaC5Y4f7XD6Ec5DP7aLk3lSUQj5PzvXSuoHo/8N3SfZ7di0+twtAFURCXDM17YzbXftg6i0ApL62g7CptK6YiuKIFp7vChti3AtH9cAM4EbgTs8YycBPzbFotoNAeJjZ9O+n9vy0ef31g0BNrvE/dAPxdrpYKqd/fL8wGqlO+c5dTmSBzj/NFMPcOZ0dZU3F/G1fmz5zNlO6mdjLbr+JnSwaTgk9YPe5zj3qygtmYho30Jtu37Zd2tppWRlZQ2NjY0dFR8fPzItLW345MmTswsLC2ued5MnT86Ojo4eFR8fP9L7evrppzv5n+t9nXfeeQ36BzRr1qykiIiI0e5rxcfHj/z0008TGuteWyvhVjh9H/CrwsTrnpfSEEw1bPnUdyyY5cMYWPGwTSPd/3rokNLw9yx29ePw9tJY85z9mvOyLaQVcr0G1jwbOJ7/MyT2dva3u7JaOrvCgQZdZS0fHVKg3x99r5GQ7bif3BVKE5sh+0dRWiIdBzkWxMLltf9tKkF59dVXV5566qlF69evjzrmmGMG3HTTTd0efvjhmnLJ06ZN2/LQQw/l1nZuY6wjPT29YuvWrYvqmlddXY0xhshIp2RKRUUF0dHh95Oq7/x9SUjLh4hku7b7+L+AHp6X0hB2Lgws1hQs5mPDmzD3clh+Dyy/b+/e0235iO8JvVxZHuterb3N/Y6fgvufd/hZTHZ872y7u5Um9objfoTffBxoQnZbPip3O9vNkXqsKC2RvYn7WH4/vJUJi9tYefoG0rNnz8rx48fvWrx4cdy+Xos/Bx544MDLLrssa9SoUfvFx8ePWr58eYyIjP7HP/6R3qtXryHZ2dlDAf71r3+l9ezZc0jHjh1H/OY3v+mXk5NTozCCzW8N1OZ2WezaXgWs9Hx1v1YGOU8JB3+XCwR3u+S86Gzv2Esvl0+Dtp6QMZ7yCE+X0z1bfHug+LP2OWc71pU95e+uCWX5qA13xoubUA3eFKWt01DxUVUGC2+w8VmLb4FtX9d9Thtn9erV0Z9//nly7969y/b1WoIxc+bM1KeeeiqnqKhoXv/+/csB3n333ZQff/xx+a+//rrkf//7X9Kdd96Z9fLLL6/ZsmXLwh49epSdfvrpfdzXcM/fN3dRf0KKD2NMkms7whgT6fnqfoVdUldEUkXkLREpFpF1IhKkuAKIyAcistv1KheRxa7j2SIyW0RKROQXEWmdJRWDio8C3/2KXZD7obMfqv5FGFRUEGj5iIhiW9x4ZyznJd+Tdi6Ar8+EdwfCyiec8RF3O9veoFOw//C87xEZF34xqmD1O6ISW2xlR0VpcsKt9VG1B6pc6ei7foFq1zN23l9847TaEWeddVa/hISEkf369RvWuXPnynvvvdfHxfLkk092TUpKGpGUlDSiU6dOw/3P9R5LSkoa8a9//av+0e4e8vLyot3XSkpKGrFr166aZ++UKVN2jBkzZk90dDQxMTEG4LrrrtuSkZFRlZiYaKZPn546ZcqUHYcffnhJXFyceeihhzYtWLAg4ddff+3gvYZ7fkPX2dyEW2QsS0Q6+Y11EhH/OJDaeBQoBzKAs4HHRSSgtKQx5nhjTKL3BXyLb2zJK8B8oDM2AHamiKTXYx37nspSW9PDH3/Lx8Z3ff+RlKzz/UcTJq++CikphqNvnUFVdYStbRFrv2Vb445yJm54w+l3Ygx8fQasfw2KVlBTsTShN/Q+1+k+W1HopMe6rR6pY2zgXDgEs3wk9dOGWkr7JdlVxWD3amvR8CfnFeteeTsTijx/gwV+oQX5c21BwHbIyy+/vKq4uHj+rFmzfl29enXsli1bfGIc//jHP24pKipaUFRUtGDnzp0L/c/1HisqKlpw1VVXBW1o5A4iXblyZYdgc9LT0yvc1yoqKlqQnJxcowh79OgR8E+9d+/eNWNbtmzp0KtXr5pfgI4dO1anpKRUrVu3LjrY/NZCuNkubwPd/ca6A28FTg1ERBKAycDNxpjdntLt/wPOreO8bOAI4AXP/gBgFHCLMabUGPMG1j00Ocz7aBnkfe2IihiXbvKP+Vg/w3ffVEPx2nq/3T//CSUlwmdLj+aL5Ufa7qSeip1F0YMc90bFLlu9FKA017fHikTYINWDn/H0axnjHPPGffjEe9Sj9pzH8lFaHsv67Z4wojqCTY2B/Pzw30JRWhVRCY5F0FTBvCsdC6MxsPg2+PYs+4GlbAes8lgm/cUHWDdMRaPETrZKTjzxxN1TpkzZ8Ze//MX/GbbXlJSUzPe+vC6T+hKsUJt7rGvXruXr1q2L8e7v2rUroqCgILJXr14Vwea3FsKt8zHAGOOOAcEYs1hEwu3VPQCoNMascI0tBOpqWnAe8JUxJsezvz+wxhjj/kta6BkPQESmAlMBMjIymDNnTpjL9WX37t0NPjcYPYpm4o1m2Bx5AN14HwBTtpMvZs8GESKrd3PYlg8C1OGib98kP9a3qn1CxSo6lc2jsMMQtlfvT9HuaNLS7N9BVRUsXToWr85ctH4YI0cWsNBzP7uLi1nHIfTCfnLKnfcsK9Z0pnPp13idJoXRg1mYdj/VEsPunyL55JMVJO8+lxuO+IK4DnvYsPAtVud0Zfj2j/Cax5ZsSWR7uN8zU82oPR0ZdPVScndm8eQFU5lwSjFrQ5xfXi5cdtkoVq1K5M9/XsnEiUED1u39NfLPriWh99Z6Cef+suVIsu3nLlj5GBtzN7MrejAZpZ/Qucw3/qtg9UcsKDyRYTvmkOoZMwiCgT1bWPfBxaxNvoj2yvXXX7+1b9++Q7/77ru4Qw45pHRfr6c+nHXWWfnnn39+n/POO2/HyJEj91x++eVZw4cPLx44cGCrs3a4CVd85IlIP2NMTdCBiPQDdtRyjptEYJffWCFQV+Wc84A7/a5TGOQ6WcFONsY8BTwFMGbMGDNu3Lgwl+vLnDlzaOi5QZn/HnjcuN36HwFL5kBVCUI14w4fY7NB1k6HLYFtE4Zlx8N+nrWUF8KimzxVRw0FxR058dqVbC1IZ/pzpZz9uzhWrfLEe3hYvGEonbIKGHfIuJp76zXoD/DZdAAyo9eROW4cLPocPF6gjn0mMKT3BB54AB5+GAoLAQaQXLqAa066jx4JW+kx9giY6VhKhhx5gbWwhMk7n55J7k77Y3x69kVMvX4RvfqOCzp31ixY4ZGxs2cP4IEHBoS8bqP/7FoQem+tl7Dur/oI+K4C1r0CQPfitwhlbE6pWsm4Iw6Fd5z+UDLkJlhiyzL1KnmDXkf9vf79kdoImZmZlaeddtqOW2+9NfOjjz5aXdf8M888s7+7Yd5hhx2265NPPqnzvGDk5eVFx8fH+/SYeOyxx9b+/ve/Lwjn/FNPPbVoxYoVuVOmTOm7a9euqFGjRu2eMWPGmoaspSURrvj4L/CGiNwIrAH6YouN/SfM83cD/hWekrH9YYIiIocDXbHFzRp8nRaJO8U2Js3WvigtsfvlBVZ8uF0ucd1sMCc4HV/z58Kck2yWiocPFx3H1gLrxnnigU2c/bt+LFvm+9aLNwyFhHW+g6kHgERa827hUitqXCm0ZQkHMHo05OT4nrZo/TBnLYWLodLTPTYuq17CA2B94ZCa7WWbBlMdXxLSJzh7trOdG9rooSitm4hIOOQFW+l0fZCSSv2nQe77tidS1R5bN8j7/yAyHob8DTZ/ZLPcRtwN8e2nMsKmTZsW+4+99NJLNRH3b7zxRk59zm0oJ510UlF1dfXcUMd//PHHX/3HjDEB86+99tq8a6+9Ni/YNYLNbw2EG/NxNzAduA/4CbjHs393bSe5WAFEiYirbB/DgaW1nPM74E1jjKvwA0uBPp4eM+Fep+WxJ4j48FJRYP2zm11diPe7ytn2Zrz8cKGP8CB1DBsKHAvA/F+7UV1lAsTHko1DqIrx+/QTnegUHcPYUuQ7nd/neRsPDRAeAPmlnpTbyt2w7B7nQH3iPTxsKHDSakvKEsjJD902yC0+tm/3tewEo6AAvvnGuqAUpVUREQWHvgR9zreB4p0PgmF3wPEL4YDHIO1QZ+5qV5fqlCH23MNegZNXQJ/faWdmpUUR1m+jMabaGHOvMWY/Y0yCMWaQMeY+Y8LL4TLGFANvAreLSIKIHAZMBF4MNl9E4oAzgOf8rrMCWADcIiKxIjIJGAa8Ec46WgwBlg9XIlH5Thvo6e3mmjwQuk1wjhetgj15Ng0WbEbJ4TNgwo9s7PS3mmnFexJYMT8nQHyUlsezZnuQB7v7n9iG12HPNrsdlUTONucTU7orPjZ/j8u64TENA9DZVVwsTDbs8P1UtmRV8DTb/HxYsMDZNwbygn4esJSXR3DggXD44fCnP9V7WYqy74mIhoP/C2fugQnfw5CboJPnw4K7kN+md51t74eJxD7aF0ZpkYTrdkFEjgHOBLoYY04WkTFAsjHm8zAv8Ses+2YbNlZkmjFmqYgcAXzgSav1cipQAMz2v4hnDc9hIxLWA6cbY2p5/OxbPv0UnnwSzj8fTjjBM+gWH7HpEJ3i7JcXQIXL2BPf0/4D8VKc41uCPHUM9PwtABs3+f44532ximXLeuPP4jXZ9D/MbzD9UFjp6VjrrveROoqcnxyNOno0fOgpPZK/p4fjrnHTEMtHnq/YWLI0glMmBs778ksn6N/L5s2QGSLpe8GCFFZ6QlHeesv+LBSlzeAW+u6/wxpLZjPxsoxutvc6q3W6GRRfwq3zcRnwOLai6VjPcCm+waC1YozJN8ac6rGc9DTGvOwZ/8pPeGCMecUY08sY/8cMGGNyjDHjjDFxxpiBxphP/ee0FIyB886DmTPht7/1BmpSt+XD7U6J7Wpbusd5YmpNpa84cLWZ3+Bq3QLw8w+l/BKkJ9WilUGsCmmuDBpvrQ+A1NGsdWX3jhrlbOfvSoSTfrFt7yM9befju0PnAwOvXwcbtvqUkWFJiDp9s4PI0S1bAse8/PSTc928PCgpqffSFKXl0mmEdcf409ziQ1HqSbhOwL8ARxtj7ga8rpZfgNCOeYXiYvupHOxD7+23geoKp56HRFirhzvmo7zAV3zEeeIq3H1ONs1yttOPqNnc6AS6A/D2F8MoLg5c1+KlQWrhJGT7lk33kjraJ97DLT527gST2A/GPAyTNsLYt2HCzxAZG3idWqiqgk2bfQuS1Ud8eL/HwZg711fUrFsXYqKitEYiO0BqEKNDuNWFlSajX79++8+aNSsJ4Morr8ycOHFioBm6gVx33XVdp0yZ0qpTl8J1uyQB3s/VXmtENLZiqRKC7X418V5+GX53his7uUOqjWh3u10qCmzRIC+xLvHhdbeYSud4uo3VqKgItACs3ZZds90pIZ+dxbYCwOJgsdwi1vqx0S+VL3WMj/gYOBASEqywqqqCoiJITsZab7oH8ZOEwZYtgcGgv/xi78ndoHH79uBrD2X5yM2FtWt9jGrk5MCgQcHnK0qrJO1g2P6tsx/fHWJSQ89vClqoK+SJJ55IfeSRRzLWrFkTm5CQUDVo0KDSG2+8cfOECRN21312cCZPnpydlZVVHqojrpdVq1Y1SiLErFmzki644ILe7s64d999dy323tZBuJaPL4Hr/Mb+TPCYDMWDfyDkZ5/Bto2uEuoxnnYB/m6XUtdH+bhu9muwip8dB9eUOc/NDYyFcHPKqP8RIfYJv2pVCPdD+qG++1FJVCf087EW9OoFqa7/a41RZdTfXQRWeKxc6Tv2xReB8yC05ePTIA65YFk7itKqSfML8O6oVg+AW2+9NeOGG27occ0112zesmXLwo0bNy6++OKLt7355pspTfm+FXWl3ylA+OLjMmCSiOQASSLyKzYb5cqmWlhbwN/yUVUFr8909eKrER8pzlg4bhcvtbhc/Bndey79u9qnuTGwNJgmT/MTH6mj2bI1gnKPfatzZ0hKanzxsX598HF/14vb5dLP9e0IZfn4JEjvvrbudiltVbUblUYhzbficU0mTDtmx44dkffcc0/mfffdt/53v/tdQXJycnVMTIw566yzCp988smNpaWl8oc//KFHly5dhnXp0mXYH/7whx6lpaUC1tKQkZEx7JZbbslITU0dnp6ePuzf//53Z4D77rsv7Z133kl9/PHHu8bHx4/8zW9+0w8gKytr6I033th1wIABgxMSEkZVVFSQlZU19O23365JNSorK5MTTzyxT0JCwsjBgwcP+u677+K8x0Rk9JIlS2qCdyZPnpz95z//OXPXrl0Rp59+en9vobL4+PiROTk50f5unJdeeqljv3799k9KShpx4IEHDpw3b16N7zsrK2vo3/72t4wBAwYMTkpKGnHiiSf2KSkp2ef12OsUHyISAQzC9lg5AzgLW4PjQGNMqzf9NCX+4gPglTddT25vXxf/Oh+lfgGnEEJ8OMGmdYmPwVnLGNrD8VkEdb2kjvJtBucXbJqd7RluBssH+IqPwkJ47z1n///+z9kOJj6MCS4+2rLlY+pUSEyEv/51X69EaVbiuzsB6aDBpsDs2bMTysvLI84999ydwY5ff/313ebOnZswf/78ZQsWLFg2f/78hOuuu66b9/iOHTuiCwsLIzdv3rzokUceWXfdddf1zMvLi7z66qu3T5w4MX/atGlbSkpK5n/++ec1Vb/feOON1Pfff39lfn7+/Ojo6ID3/PTTT1NOP/30nfn5+QtOP/30/NNPP71fWVlZrSIgOTm5eubMmSvT09MrvD1ksrOzfUwrixYtirnwwgv73HvvvRu2b9++8Nhjjy049dRT++3Zs6fm2m+99Vbqxx9/vHLVqlWLly9fHvfII480uEtvY1Gn+PDU8njH08jtR2PM68aY78Ot8dGeCSY+vvk5jXV5Pe1OKLdLMMtHolOEq4YuwS0fWUGKzQ/OWsawnk7TKX/x8e678OY7sZgUV/CaX7Bpc4iPIU6h0xrxUVgIEyY4wiEhAU47zZkXzO2yeDFs3Ro4Xh/xUVpqY1paA5s2wdNPQ3U1/OtfsGdP3ecobYiep9uvkfGQMX7frqUFkJeXF5WSklIZTASAFQo33HDD5qysrMrMzMzKm266KXfmzJmdvcejoqLMvffemxsTE2OmTJlSGBcXV71o0aJaI+kvvvjirf369asI1dZ+//33Lzn//PN3xsTEmFtuuWVreXm5zJ49O2GvbhR48cUXU8ePH184adKkXTExMea2227bumfPnohPP/20JuBt2rRpW7OzsysyMjKqjj322MIFCxbE1XbN5iDsmA8RqX/lqLbKpvfhy9Ngzon29cOFUBJoeggmPgBe/f5Mu+EVH+6A05KNtmIoQEQH51h0km82Snx3WwPEg/sBfsopvu+XEr+TrilbfCwfi1zNL995x54zeTL89e3nbepex/2h+8Sg4qOTSyt5xceOHfDYY4HunJUr4euva49Hca/9+OOd7SVLYNcuOO44+OEHZ/yhh6CPq/TJli2B13dbPQ52/ebW5XaprLRCbMoUK7JSUuD1IJWtmwJjav8+1caXXzrbVVUEFJdT2jjD/wGHvgITfnDixNox6enplQUFBVGh4i/y8vI69O3bt6ZNfZ8+fcq3bdtWo1Q6duzoI1zi4uKqi4qKan1eurvMBiMzM7MmQSMyMpKMjIyKDRs2BFdH9SA3Nze6R48ePtfu1q1bufvamZmZNWuLj4+vLi4ujvS/TnMTrvhYB3wgIs+JyB0icrv31ZSLa5EUb4CvTrNZIbnv29fqZ2DuXwKmusWH+wH4yrcen0GwmI/iHGc7tqvNQvHidr2kH+5zzG35OPxw6NTJMUwN7rECiU1j6FgnT3bRIudBN8PVRubeJwcwUwrghMUQFe8jPnp7PIzBLB/nnQeXXAJHHAHbtjnvMXy4HXvkEULiFh/HHgsRnt/KVavgsMPg+++d4489Bn/4g409ifNo99JSK1LcfPyxs/3730Ok509ty5bQcRHbtsEhh1ghNmOGtR5UV8NVV1lR0pR8+SWkpcGwYbYcfEPOd+MWl0o7ICoOss+0ZdUVxo8fXxwdHV09ffr0TsGOp6enl69evbomxmLt2rUdunTpElakaKj29SJS60eH3NzcmhoHVVVVbN26NbpHjx4VALGxsdXFxcU1z2O3EKrrupmZmRUbNmyouXZ1dTWbN2/u4L12SyVc8REHvI1Ns+0O9PC86tc9rC2w7G6oLgscz33ftzgXvtkuF55XQIznV33h+hGs2Nw/uNvF7c2Kc1k6wGa3eOky1ueQW3z06AGjRjk/2sFHHAST8+h9yt/o2NGObd9u01mNCcwiOf/CWJb/Yv/A6nK77PR4VL/5xtl/1FMo9Y47nAe9O17DH7f46N/fCSY1xjfu45FHYNo0uy0C3Vwf8NxxHytW+Ga6HH+8/Z54CRbgunEjjB0LP/8ceGzDBmsNaSrKymwF3Px8e7//+1/9r+EvPhYubJy1NQaVlfbndsQRLWtdStulc+fOVddee23u1Vdf3fPFF19MKSoqiigrK5MZM2YkX3zxxd0nTZqUf/fdd3fLzc2N2rx5c9Rdd93VbfLkyWF1ae/SpUvF2rVrg1R2q52lS5fGP//88ykVFRXccccdGR06dDDjx48vBhg0aFDp888/n1pZWcnMmTOTf/rpp5pA1czMzMrCwsKoHTt2BLVWnHPOOfmzZ8/u+M477ySVlZXJrbfemtGhQwdz9NFHNziduDkIV3xcZYw5P8jrD026upZGyUZY7WrkO+oBp/R5VSlsneMzffs2pwxKn8JrOO44R8C+/sNvXW6XZCCImo7rRmGhDSD8wx9ga6erodMo6HY89P69z1T3A7x7d2sx8DLaE8YREQHjXe7gTz+FzZtj2bTJ921374ZJk6xwCCfgtLjYVb0VKz4WLIA3XB133NdxU17uxGZERNgy6UP9MgU7dIBnn7WWFTddXdrMHffx979biwXAAQfk07OnTRH24h/3sWaNtRb9+quzjquust9zL17LTVERPPGErxvITVUV3HMP3HmnFRXh8PDDdg1eQgXghiIvL9DN0pIsH888Y79nX38Nt922r1ejtBduu+22rXfeeeeGf/7zn926dOkyPCsra9hjjz3WZfLkyTvvvvvuzcOHDy8ePnz44GHDhg0eOnRoyd13311LuUKHadOmbV+5cmVcUlLSiKOPPjpIMF5wjj766IIZM2akpqSkjHzttdc6v/baa6tjYmIMwIMPPrj+448/TunYsePI6dOndz7mmGNqAmVHjhy55+STT87v27fv0KSkpBE5OTk+rprhw4eXPfnkk2uvvPLKnmlpacM/+OCDlLfffntlbGxsA524zUOtRcY8cR5vAhkish6YZIxZ0BwLa5Esvdtp+JZ2CAy83Laz/vVBO5b7HmQeZ7erytm+YRNgfRVpfM8ZE3fyzjv2yT3jhzO4MdZzLYmA6I5O5VMPC9eP5vQLrPsBoHPn/tx7b2Atn4oK34dvVhZccQWsXm1dE7//vXPsmGM8lVaxcRGDB6fUHNt/f/sQLC21D+KXX/a1Engf4P7iw1+87Nhh+9i44xdycqwgiPCTu5s2OfO6dbNFxcaMcYRLly7w5pu+YspLMMvHmjUwfbozfu65OUAq2dmOhccd91FeDhMnOmPR0fa+Tz/dioDnn7eC4vPP7cPzkkvsgz0uDpYv9xU1YM/1ZpskJ8Of/xy4bjd5edZC5Ka2iq3B+PrrwLGFC+33NYSFuNmoqIB//MPZnz9/361FaX9MmzYtf9q0aUHD4p977rkNOMUzazjppJOK3AW9ADZt2lQTMDd06NCyX375ZVmo48HG7r///loLko0dO7aktqJkr7/+eo573/965513XsF5551XEOxc/7XVtZbmoi7Lx33YzrNDgZme/fZJyUbfltVDb7X/2bNOdMY2vec8Sef+me0FTiBzWtJ2Th67hJho65pZtH44v+S4np7uuA/gxa/O4eALr6sRHhC65Lg74DIjw1oKUlLsQ/jppyHWFaN9zDHO9pw5MG+e877nnAN/cxrj8q9/Oe3q09JsGicEio/cIL/K/g/Q8vLgD1X3p3yva+RPf7KptGedBT/+GFx4gK/lwys+7r7bqZY6fjwMHWqDQbxWG/C1fNxzj/N9jY21wbenn+6s59RTnbm/+Y1jUSgtDe4ecQuBz8NouXjrrYHxKsG+n7Xh73IBKwDre52m4MUXfcVeTo61rCmK0r6pS3wMBm4wxiwDbgLabzTTsn/6Wj26ep7i6WMhyvNULl4Lu36FVU9TveIpdhTVZG6RlrSdpOplnDDio5qx199NZ/lyuOACuOvNqyivtNa0WfNO5HdPPs+eMt9A6FC1PNzj3euIwunXD3p6kmSKimD2bKfJ3NixcO65zqfl5cud89wP77osH6EIluYaTHwkJ1sLwksvBVoW3Pi7Xdavh+eec8bcQiqY+Pj1V1+rw9//7pttA3Dppc62f+B8sDoibvfHvHnB111cbB/KJ59s3RH+1Nfy4RYfHVxte/a166Wy0n5P/XH/XimK0j6pS3xEGWP7NBtjyoAgHcnaPlHVhTajxcuQW5wndGQHR4gALL8H5v6ZncWdqDY2PqhjfAHRUZWwcwFTDnq5ZupjT3bggAPgv/+Fm164lFP+9T/mrh3FWY++jDH2R+Ou5BkqFiDYAzwUIr7Wj8pK+z6xsdbdkZXlGxfixZvpArVbPrwZKF66uBroBov7qM/a/fF3uzz4oCMQDj8cjjzSOe4WMevWWRfQ1KnUVG8dMya4i+TII607Khhz5vgKEmN8xceGDYEl9isq4IADbHbQrFlObMpAV4vG+oiPwkIbXwPWpfXb3zrH9nVw5yuvWNefP0Gr6yqK0q6oS3zEisgL3heQ4N73jLV5Motn2YBSsC2sux3rO8HtelnzLFTtYXuRU0AuLcmTc7v9W04c+R5xHWxjlS1bxKfr7EeLjuOAm3+iaE8yAL16lPHNN86n2cLC4EWv6mP5AF/x4eWQQ5z3OfvswONuy4F/nQ+35ePSS63LB2wH3HPPdY4FEx/umJL6ig9/y8ebbzr711/vG+/gb/l47jnHYhAZCf/5j5OO60YEbr7Z2b/vPl/LkTvwNC8vsOiav/Vj7tzAT/5jx/pm02zeHH69j2+/dQTMiBE2o8TLvrR8VFTAXXc5++npzraKD0VR6hIfdwGrXa+/++0H+VzTxqgqJ6v4bWd/4BWBUXyZJwSctr3EKTOalugRHwVLSIwt5sQRvnmnbuuA1+IR16GEt1/bQZcuvoIimOulvuLjqKMCx8a6MncnT6YmLdiL++GdkOB0m92zx/fT7YgR9qF+//3w0Ue+xcCa0vLx3XdObEFycqDA6t7dCXbNzYWbbnKOXXONrUcSiilT7PXnzrVZMO5ru10vwQp7+YsPd7O8Qw6x37svvrApxslWc1JeHl7lWGPg/fed/bFjbZ0QL/vS8vHww072UHKyr/tFxYeiKLWKD2PMbXW9mmuh+4z1rxNT7REPsV2h15mBc+K62RRYF9u7OE2Aaywf2I+zFx/lOPonTrR1KR7565uIODU+np16PiMOtP6NusRHfR/gaWkwcqTvmFt8dOwYWCXVLT5EfF0v7kDYrCybKnvFFfZ93O6axhYfbsuHO4jxuOMcceQlOtr3++h1bXTt6itEQnHwwdaSA/UTH3P9kpPc4uPII33FWWams11XsOjnn8Nll430Kd42dqz93nu18a+/7psy67m5NpDWy803w6GunoUqPhRFCbfOR/vEGPj1AWd/wCU2xiMYPSY52z1/y/Yo5wmV1sm3pOZRQz7n28dv5pNPrKugY0e45Oyl/O/KUzhu2Ac8dcFFTBn7MUTaNBX3Q7kxLB/g+wCNjvatwAqBrhe3+ABf8eEOJPXvK9OU4qNLl+CppCedFHx+sODVm2+2lpz6cNRRzvv++KNT4yQcy4c7e8kdzwO+lpza4j7eeQeOPhqWLu1YM9ali11XYiL09VQeaIoy659+aq0/7vvw55prHPfgoEFw+eXWsuMVhOvXt56eOYqiNA211vlo79zwlw0894Inn1IEYroErQVmuREqLmP4gFxeebs7279xJqYF6R94yOidcIBroEMKJ416j5NGeVwycfvVHHILCu/DeskSmybaqZO1nASbWxvHHWfPBys84uN9jx9/vPXT5+XZUuZuEQG+4sON+9M7+IqWDRtsLID3IbRzp+NeiI62acL1ISrKrtFbzh3sj8k/Y8W9lq++cvZ794YLL6zfe4JjOZo3zz7gZ8+2KbnBHvRr19r79MbJuC0f/fv7zg3H8mGMtSp4Y0I6dICLLoIbbnDcNsOGOeJg0SLHYhOKxx+3lpRbbvFt7OfPli3WIlZaajOR5s4NFJtz5thMJS+PPOL8vAcMcKwey5bBQQfVvi5FUdouavmohcL1S9lckGlfO7uxeUskmzcT4iVs3t6RD78dxBPPJPlkOaR3CdI7KMZPkbhLrINPE7lglo+pU2265kMP+cYHBOtoG4xx46xrZPDgQu4LUr2lQweYORPOOMO+j38WSzDxkZISKGJiY51P9NXVvpYOtxAYMSKwAFk4uK0FYOMogok9CLTe3HGHb2pqfQjmenEHkrrX4C2sZUzt4iMcy8dPPznZLTExVfzyi33Au4WLO36lrriPX3+1dVVmzrQp37XxzjtOufytW209FP8qrre7uj1NmWJro3gZ7OoOoK4XRdl7RGT0kiVLYgDOOuusntdcc02DugrGx8ePXLZsWbNms6r4CIWprnF71Jeff/ZtKpfWLTFwUoD4SPHdd/V18Y/5qKgI3oOkV6/AQNFQiNig0Ecfnc+BBwafM3YsvPaajUvxJ5j4CCV83FYTt4vG3VPGnRZbH9xxHxDa5QK+bo6hQ20hs4biFh8ff2wFoLfQWUyMb8yMN+5j+3bHRZOYGLj2cCwf7rog48dvC7BIQf3Eh7so2o8/hq4lA1Z8uPn+e+tS8VJd7Zv9465sCr4pyyo+lKYkKytraGxs7Kj4+PiRaWlpwydPnpxdWFhY87ybPHlydnR09Kj4+PiR3tfTTz/dyf9c7+u8887rGfrdQjNr1qykiIiI0fHx8SMTEhJGZmdnD/n3v//due4z68/LL7+8/t57760zUf/AAw8ceP/99/s8gEpKSuYPHjy4PNQ5TUFY4kMsF4nI5yKyyDM2VkTOaNrl7UMkgrv+M55Nq/KY88Lf2bSJWl/uLqqLFvmJj8wgT+rYdN/9AMuHI2D93S4rVjj1JTp1gtNOs3Utausc29gEEx/+LhcvoeI+5sxxthsqPvwtH7WJj0mTrIspO9v2G2mIpcXLYYc5Vp5Vq2yqrpf99rO1PLx44z7cVo9+/QLjVeqyfBQUwKuvOvsnnxz8/4x/xkttabs//eS7H6qpXVERfPZZ4PiTT8Lrr9vtNWugxGaR06VLoKtOxYfSnLz66qsrS0pK5s+bN2/pkiVL4m+66Saf/xbTpk3bUlJSMt/7uuiii3b6n+t9vfDCC0HaUYZHenp6RUlJyfyioqL5d9xxx8arrroqe+7cuQGfbCv8qxi2ccL993s7cAHwFOBVgBuBvzbFoloKKSmQ2Tcd0+NQMjOp9XXEEU6diFWrfEtKp3UPEszgb/mITvHdjwvtdnFnlxxyiO2D8tVXtT94G5tOnQLHQlk+3O4Or/jwL451+OENW4fbetCzZ+0xC0lJNmV27VpfcdAQYmNt+Xcv7kqpgwf7xlmEEh/+uMVHMMvH9OmO22PYMBg0aFfgJOz32xv/UVcFWn8Lmr91w8tHHzkF2YYN87UaeavKuuuKuAWQFxUfe0lFhY34ffZZawKbMaNl1NBv4fTs2bNy/PjxuxYvXhxX9+ymIyIignPPPbcgKSmpctGiRXEPPfRQ51GjRu13wQUX9EhJSRlx1VVXZZaWlsrUqVO7d+vWbWjnzp2Hn3XWWT13795d8zHl5ptvzkhPTx/WpUuXYQ8++KCPBWXy5MnZf/7zn2s+Ak6fPj1lv/32G5yYmDiyR48eQ2bOnJl82WWXZc2dOzfx+uuv7+m26LjdNzt27IicNGlSdqdOnYZnZmYOvfbaa7tVeXpWPPTQQ51Hjx49cOrUqd2Tk5NHZGVlDZ0xY0Zyg74fYc77PXCSMeZVvPmisBboE/KMdkZsrFOl0r8VfFr3rhDh506ry+3iivlIS3NiEwoK7APUi38H2OaioW4Xr/j4+mvf4ljewmT1xR03ccopzdtIzV163Z3qO3iwffh6xeiKFbZ/iztDxD/eA3wtR/6WD2OslcHLH/8Y+l5FwnO9lJUFFiKbPTuw1wz4ipKJE21wqpeffrLrW+xqXxXs97JfPyf4dOPG4O+jBGHrVtsOODvb+vv+8AeYNs0G1fTsaYNv3GZExYfVq1dHf/7558m9e/cOs89001BVVcULL7yQUlRUFDlq1KgSgEWLFiX06dOnLC8vb8Hf//73zZdeemn3VatWxS5YsGDZqlWrFm/ZsqXDX//610yAmTNnJj/22GNd33///RWrVq1a8tlnn4V86M+ePTt+2rRp2f/4xz82FhYWzv/yyy9/7du3b/nDDz+8afTo0bv/8Y9/rA9l0bnwwgt77Nq1K3LNmjWLZ8+e/euMGTM6P/TQQzUPrIULFyYMHDhwT35+/oI///nPWy699NLsau8/83oQrviIBLz/Xr3iI9E1phD80x5AWpdISPL7qFtXwKnL8hER4et6+fBDZ7u2T/pNyd66Xdz/K8eNa/g6pkyxFohTTvF9IDYHw4f7VhT1MmiQFaPuT/o//1x7sCkEWj6MsQLtjTdg9GhH0MbHB69C6782L6HEx8KFgf1qKirs71duLvzlL3DvvVZYveeqizdxol1/R0+mb16etfTVZfmIjvYtI9/YacBtkjlzrB/v1luDWzmqquwvyPjxcPHFtqGOAsBZZ53VLyEhYWS/fv2Gde7cufLee+/1+QY++eSTXZOSkkYkJSWN6NSp03D/c73HkpKSRvzrX/8KEcZeN3l5edFJSUkjUlNTR/z973/PfOyxx9YOHz68DCA9Pb38xhtv3BYdHU18fLx5+eWX0x5++OENGRkZVZ06daq+4YYbNr/99tupAK+99lrqlClTth9wwAF7kpOTq++6666QZq+nn346/YwzztgxadKkXZGRkfTu3bti5MiRdVb9qaysZNasWan33HPPpk6dOlUPHDiw/JJLLtnyyiuv1FhZMjMzy6+66qrtUVFR/OlPf9qRl5cXvXHjxnpnzoZ7wvvA/SJyBdgYEOAO4N1az2pnDBvm65MHKxw6dQKSBkCh679tjF/MUWQcRERDtedpEOsbjdi9u/Wpg1M5Elqv5aMxgk3B1uh46aWGn7+3XHqpb9YOOFkdhx7qPJDffbdu8ZGYaF1DRUVOldNzzvEVm2CzUrwP/lCEIz7cLpeICMcS9d//2sJr3vU+9JBNFwb7ezhqlLWuHHCA9QKADVaty/IBVpB5RdSSJYH1ZcD+brz2Wg+GDoXOTRKa10r48kuYMMHxd4HNRz/qKKtAly2z9fW9PPmk7Vr4wgvNawJsobz88surTj311KL33nsv8fzzz++zZcuWqLS0tCrv8T/+8Y9bHnrooaAPcO+5db1HfHx8TbnGhQsXLu3fv39A0GZ6enrF1q1bgzY76NatW43837x5c9SePXsiDj744EHuOVVVVQKwZcuW6FGjRtU05Aj2Xl42bdoUPWHChMK61u/P5s2boyorK8V97d69e5dv3bq1JmUzPT29Zs1JSUnVALt27YoE6qV8w7V8XAl0AwqBjliLRy/aeMxHfQlWortzZ09gY7LrI19UYmAmjQjEe4I7JAoSfCtuBSvAFRlpPxTtC+ojPnr0cFwQW7bYuhzeOAiR4NaD1sKkSb4Wi6goJ55jkqvu3Jtv1u12Ad9rffCBr/CIi7Pp0cFSo/0JR3y4g03POcfZ/ugjX6HkzoBxu7bcWVKzZzv3FxHhm1brZsQIZ9vdjdfLunVw7LHwxBN9ufLK4NdoF+zYYX+BvMKja1dbQGX9equ2n34avvnGKjj3L9r06c0bed4KOPHEE3dPmTJlx1/+8pcwqyCFjzsotTYxEAoRqQkH79q1a2VsbGz1okWLlhYVFS3wvkpKSuYDZGRkVGzYsKHGf79q1aqQqbFZWVkVq1evDpqu6X5Pf7p161YZFRVlVq5cWXPtnJycDhkZGY0eDRuW+DDG7DLGTMIKjoOBvsaYScYYrVPoIpipuabeQ9IAZzAmPXAiwMh7odNIGPHPADdMsOJhAwaEn1rb2NRHfERF+Yqnl16y1mKwD8lgwauthehoa+32MmCAE9cwbpwTy7J+vRPjkJTk28/Hjdt19eyzzva4cTZN+f77w6tNMmSIk82zYgU+DQy9uMXHhRcGCtzIyMDfL3fatVt8vPqqk1XTv39gXRgvRx/tbH/8cWAmzmefOc/bYNk17Yb//tcp4NOli7Vw/N//Bf7w99/ful3c1fL+9S/nD0wB4Prrr9/6zTffJH/33Xf7NOi0NiIjIznzzDO3X3LJJT02bdoUBbB27droN954IxngjDPOyJ8xY0ba3LlzY4uKiiJuuummEI5uuOiii/JmzJjR+Z133kmqqqpi7dq10fPnz48FSE9Pr1yzZk3QJ0dUVBQnnHDCzuuuuy5r586dEStWrOjw6KOPZpx55pk7Gvt+w021jRCRCCAPmAts8+wrLrKyAh/KNeKjo+ujYHyI35kep8Hx82BQ4Ee+YJaPfeVygcD7jIwM/UAFX9eLu8nY3rhcWgoXX+xYLM5wJZ936BDYIweCp9l6cVs+Pv/c2f7DH2r//voTF2eFEAQGQION4/AWRYuIsK4U/7X+979WoAzyGIEHDfKNz3GLj4ICZztU7BPYyrBeV8rWrYEBr+46IZs2+V63LrZsaSPJH9XVvgVd/vGPwLxlNyLw7387f5Tr1lmzmVJDZmZm5Wmnnbbj1ltvDfnAdnPmmWf2d9f5OOaYY/o29RoBHn300Y19+vQpO+iggwYlJiaOPOqoowYsX748FuCMM87YNXXq1K0TJkwY2Ldv3yHjx48PGbI9fvz4kkcffTTnmmuu6ZGcnDzyyCOPHLhmzZoOAH/5y1+2zpo1q1NycvKI3//+9wFPlv/85z/r4+Pjq/v06TN07Nix+02ePDn/8ssv3x74LnuHmDB6d4vteBZsYiWQC7wJ3GKMabEBqGPGjDE/B6vMFQZz5sxhXJhRkePH+wZTTprkafVuDHx7NmydDQc+Bd1Prtca3nnHlvB2c/vtvu3eG0J97s1NdbW1aHh/fbKyai9QdcEF9mHmz1tvBd5XY9LQ+6sv27db68bIkb7CItjP7YwzbPG2YFx9tf3g6s+WLYHl5+u6tzPPdN7nySdtVVwvX33lNBMcMsTGayxfDmPGWMvDv/9tK5+C3f/xRysqvCm8Xnr0CPy51/V7+X//58RG/fOfcO21zrGRI50UbLCeBXdTOn8WL7bP6c8+s7FQkZH2d+rk+v15NTu1/ux++MEJhunUyaowlylp40YbO3XIIfZvsIZrrnF8cmefbV0wjYCIzDXGjPEfX7hwYc7w4cMb/aGktB0WLlyYNnz48Oxgx8K1XlwGfA4cCwwCJgCfAdcC04BDgQf3dqFtAf9PfTWWDxE47GWYlFtv4QHB3S770vJRE0jroa6y7r//faAJv2dPXzN8ayYtzQnEdHPssYGN60LFe0Bg0TSw161v3xuoPe7D7XLx1jwZNMg+53JyHOEB1oJz+OGBwgMIWh23NssH2O+JF3dxvpIS36BVqL0eyK5dttjbY485QdhVVbYdQKvGXZjlsMN8hMe2bTZjaOxYK9Rmz3ad5y672yZMQEpbJtxslyuBUcYYb/TsChH5GZhrjOkrIoux7ph2j3/QaUCfkQZGoQdzu+yrNFsvnTo5bulQabZejjjC/j9cudLGH1RU2A93iUEqz7cl4uLghBOcKqBQu/gI9n087riGvXco8VFd7VsrZozrM21KSv1qrhx4oMey56IuUewWH199ZX8fEhJsGXr/UIXaxMfChcG747p77LRK3PnPfor966+dKrJLltjeOWeeaSvsJrjnltc79lFRmpVwxUcyEI/NdvESj818AdgCtNhAnuYkpOVjL/EWGvP+T4mPhz77uMRbaiqsXm23w2lol5raPjuZnnZa+OIjmOWjoeLDnVmyaJHNLrnjDmvVdz+096baq7/lIzExsIGfP1lZNk5y6VL7+/zll7YT8Y8/Bs6tTXx4U8/BxqJ43Z2//mrLXUSF+9+tpeHOL/bWx/d8aNkexMnx6qvWQHJp1QJnsLH+8YSJCKOb672M0Q+6bYFw3S4vAJ94+rscJyIXAh8Bz3uOHwv8GvJsQERSReQtESkWkXUiclYtc0eJyJcisltEtorI5a5jOSJS6jm2W0Q+DnWdfcH++/v2DEkPkdhSX/wLjfm/z77AHXQabjfd9siJJ/p+gK2P5SM5OXgtjHDo1s15BhUV2eDeTz/1FR4pKXW7SWpj9GhfY547y6Y2Jkxwtj/6yH51B5t6qU18uPsEHXKI872rqPAVJq2Ogw+2KVFg85ddQWRu8eH+nVqy2Nj0Wy9txZ+ptFnCfXxdAzwCnAk8AJwFPIqN+QCYDdSVt/AoUA5kAGcDj4vI/v6TRCQN+BB4EugM9AP8BcbJxphEz+tY/2vsS9xZBtC4RZLc4mNfu1zA91N6r177bh0tnaQkm7CQkmJjAmsTpP6Wj6OOclJ364t/mXU36elWAMycuXfp2snJTjYMhC9k3K4Xr/gIZvnYvNkpcOaPW2D06eNbW8RbPbWiwgatBnPPtFgSE+G885z9G2+sMXm6xYfb6tT76xcdpZaQ4Hu+sk/o16/f/rNmzUoCuPLKKzMnTpxYS8pS/bjuuuu6TpkypVX/1w23zke1MeYJY8xRxphBxpjfeParPMf3GGNKQ50vIgnAZOBmY8xuY8zXwP+Ac4NMvxL4yBjzkjGmzBhTZIxpVV7cKVPs1+Tk2iP160tPV1PnfRls6uWSS2wsymGH+dY5UgK54gobH3PPPbXPS0ryjYNpqMvFi9v1AjYmceVKm+b64YdW3OwthxzibI8cGXqem7FjHdHzyy/WIuNtxhgbC9nZTmGSpUttCfdLL4VHH3Wu4S8+3CLIG/dx6aU2WHbkyFZWefxPf3JMSt99Z1Omtm3zER8HHQRCNX/gGa5Y5qrz8bvf1UQHv/22/T6FkdS4VxjD3OZ61WddTzzxROqQIUMGxcfHj0xPTx82duzY/h999NFeRZr5N3ALxapVq5aedNJJey17Z82alZSRkeEj6+++++4tr7322rpQ57QGwvaKikgGcCCQBtQYWo0xQRIoAxgAVBpjVrjGFhLcWnIwsFhEvsVaPX4ALjHGuBvgvOSpMzIfuMYYE7SGo4hMBaYCZGRkMKeBzZd2795dr3OPOEJ44IFkevQoZcGCxgv8GjEihddeG0ZCQiW9ev3MnDl7f+363ps/zz9v/0cGM5m3BPb2/vYFPXuOZNmyjkRGVpOS8gNz5gTvhxXOvQ0dGk9S0kiqqoQLL1zLxImb2Lix9rTo+jJ+fBxffjmYjh0r6NVrCXPmhNdkasSIofzwgzUNTpxYifffUd++haSnF5GTY9OE3nzzV37+OZWvvrImI5F5DB68i19+OQSwCmbr1u+JikrF/quB2bO3MHr0rzz//OFAJKtXw3PP/US/fkGqre0DQv3sSkoiee21HiQnR/OXCy6g73/+Yw988AHV3btzZsdJVDKRYhKYMPdzLmYWfXFUWHGvXsw/7jgq58xh164oJk2y7aK7dNnDyy//QGRkE6uQFsStt96a8dBDD3V94IEH1k2aNGlXTEyMeeONN5LffPPNlAkTJjRZWYiKigqiG2qubE8YY+p8AadiS6rPx7pO5gMVwOwwzz8C2OI3dhEwJ8jcFUABcAAQCzwEfOM6fhg2uDUeuB4b7JpS1xpGjx5tGsrs2bMbfG5js3OnMWVljXe9lnRvTUFrvL9584w5+2xjXn219nnh3ltJSeP+zjQWixcbExtrjP1c7ryuuMKYCy9cXbN/8snGREQ4x++5x5jiYmc/MtKYigpj5sxxxkaPNuarr3yv+9FH+/qOHUL97P7+d2e9L79UbcyNNwZ+g0K8yvoPNmb9+pprzZrlHD7ggIavFfjZBPmfumDBghxjzM8t8bV9+/b5cXFxVc8888zqYMdLSkrmnn/++VvT09PL09PTy88///ytJSUlc40xP7/77ru/dunSpfxvf/vbhk6dOlWkpaWVP/jgg2uNMT/fe++9OZGRkdVRUVHVcXFxVePHjy8wxvycmZlZdsMNN2zs379/SXR0dHV5efnPmZmZZW+99davxpifr7jiitwJEybkn3DCCfnx8fFVgwYNKv7222+XetcDmMWLFy/27p922mnbL7vsstzCwsJ5MTEx1SJi4uLiquLi4qrWrl278Iorrsg95ZRTdnjnT58+fWXfvn1LExMTKw844ICiuXPnLvEey8zMLLv55ps39O/fvyQxMbHyhBNOyC8uLp7bHD8Hz+9I0GdyuDEfdwLnG2NGAsWer1MJP712NzZjxk0yEMwkVQq8ZYz5yRizB7gNOFREOgIYY74xxpQaY0qMMf/ACpVW3B2kfqSkhFdeW2m9jBxp60N53Xd7S1xcy/ydGTIEHnwwcPzAA33dLu++6zS9A5sAkpPj7PfqZTNb3DEfy5f71cDAFmpr6bhTol96WeDOO2359FpSkvLpxH1cxTf//MYnJ9/dc+6ww5pitS2X2bNnJ5SXl0ece+65QSOGrr/++m5z585NmD9//rIFCxYsmz9/fsJ1111XE3G1Y8eO6MLCwsjNmzcveuSRR9Zdd911PfPy8iKvvvrq7RMnTsyfNm3alpKSkvmff/55TcemN954I/X9999fmZ+fPz+Y5ePTTz9NOf3003fm5+cvOP300/NPP/30fmVlZbXWXkhOTq6eOXPmyvT09ApvD5ns7GyfPiuLFi2KufDCC/vce++9G7Zv377w2GOPLTj11FP77dmzp+bab731VurHH3+8ctWqVYuXL18e98gjjzRvOlQQwhUfPY0xr/uNPQ+EG9W0AogSEXec/3AgWCz7InyrqdZlJzS43ECKorQepk6F00/3HTvoIF/x4c/ChYHxHmCDaL3ZPSUlgcXGNm9uhAU3Mdu2OdsffwyFhdhc7R9/hB9/5LHoy3mD03iXkyifci7/OfS/dGcj13Afv25N8bmWW3w0ZuxZayAvLy8qJSWlMpT744033ki94YYbNmdlZVVmZmZW3nTTTbkzZ86sSQ+Iiooy9957b25MTIyZMmVKYVxcXPWiRYuCNmrzcvHFF2/t169fRWJiYtBn1v77719y/vnn74yJiTG33HLL1vLycpk9e3ZCsLn14cUXX0wdP358ode1dNttt23ds2dPxKeffloT2zJt2rSt2dnZFRkZGVXHHnts4YIFC/Z5aYxwxcc2T8wHQI6IHAL0BSLDOdkYU4wtwX67iCSIyGHARCBYLcJngUkiMkJEooGbga+NMYUi0lNEDhORDiISKyLXYGNQvgnzPhRFaUGI2AzRvp7OGVZ4QLdue0I2p/vlF/vy4q534w46dXfmhdZh+di61dmuqLBWHy9lww7gkooHOZ03mBT5LtGvvMDm486nlHjAV5BVVPhmD7U38ZGenl5ZUFAQVVFREfR4Xl5eh759+9YEU/Xp06d827ZtNUqlY8eOPsIlLi6uuqioqNbnZa9evWrt/JqZmVkTpBcZGentUrvXwSG5ubnRPXr08Ll2t27dyt3XzszMrFlbfHx8dXFxcVjP7qYkXPHxNHC4Z/sBbGrtQuCxerzXn7CxGtuAV4BpxpilInKEiNQE/xhjPgduAN7zzO2HTe0FSAIeB3YCm4DjgOONMY3ecU9RlOYhJQXmz7dpv+++awVJRISvkHBTWQnvvefsu3uuuV0v/rQG8eG2fID1uHjZ4fovl5Zmv09u4eUt+AfWOuSthNqrV/urwzN+/Pji6Ojo6unTpwftmZ2enl6+evXqmiTztWvXdujSpUtYbeMlRJXq2lrVA+Tm5tY4P6uqqti6dWt0jx49KgBiY2Ori4uLa57HbiFU13UzMzMrNmzYUHPt6upqNm/e3MF77ZZKuOLjXmPMGwDGmBewIeWjjTFhtzUzxuQbY041xiQYY3oaY172jH9ljEn0m/u4MSbLGNPJGHOyMWaDZ3ypMWaY5xqdjU39bVi3OEVRWgxJSTB5sm8NlP1dVYBiYpxGeGCronoJZfnwp6WLj8pKX4EBtjmtt0aJO83W617yWozA1/LRnl0uAJ07d6669tprc6+++uqeL774YkpRUVFEWVmZzJgxI/niiy/uPmnSpPy77767W25ubtTmzZuj7rrrrm6TJ08O60Nsly5dKtauXVvv6jhLly6Nf/7551MqKiq44447Mjp06GDGjx9fDDBo0KDS559/PrWyspKZM2cm//TTT0ne8zIzMysLCwujduzYEdRacc455+TPnj274zvvvJNUVlYmt956a0aHDh3M0Ucf3WIbvUIY4kNEIoFiEan5Zhtj1ptWVntDUZTWxahRzvb//Z/tY+LFHYDqFh+1WT5aeszH9u2B9TjKyuD9953jXrziw9/y4T2/vYsPgNtuu23rnXfeueGf//xnty5dugzPysoa9thjj3WZPHnyzrvvvnvz8OHDi4cPHz542LBhg4cOHVpy9913h/UbMm3atO0rV66MS0pKGnH00Uf3rfsMy9FHH10wY8aM1JSUlJGvvfZa59dee211TEyMAXjwwQfXf/zxxykdO3YcOX369M7HHHNMTaDsyJEj95x88sn5ffv2HZqUlDQiJyfHx1UzfPjwsieffHLtlVde2TMtLW34Bx98kPL222+vjI2NbdF51WLCqD4jIgux7o1W2ypxzJgx5uefG2Ykaa627PuCtnxv0Lbvr63f28iR4zjnHFvc84UX7AP1tNMC5+7Y4ZT637gxeBNGgI4doaCgqVZcP4L97BYuDCwKB/Db38KMGfDaa7aJHFgr0cyZVmwkJ8Nuz2fcbdus9ahnT9iwwY7Nnesr5OqLiMw1xozxH1+4cGHO8OHDg3SbURTLwoUL04YPH54d7Fi4RcZeAmaJyL+BjbgyUDwxGoqiKI1Kx46+AZfBSsV37Gi7K3vJyrIuHK+romdPa/GoqLCZI6WlhAxk3de44z0yM20XaLDxLSUlwS0fItb14k3RXbMG9uxxhEdCwt717lGUpiLcmI9pQCfgVuA/wDOe13+aZlmKoii+ZGc7/da89Onj29hOxDfuY9w4yMhw9t3ZJC0N99qOOMJpQFhSYsVFMPEBga4Xt8vloINacXdfpU0Tbm+X3iFe+7ipu6Io7YWIiMBP8e5MFy+jXc3djz4aunZ19lty3Ifb8tGli2//ptWrQ4sP/6BTjfdQWgP16e0Sje27kmmMec3TLM5bw0NRFKXJGTbMdqn10ifIx5/rrrNuh6wsG6g6Y4ZzrCVnvLgtHxkZtsGel9rEh7/lY/FiZ7+JK5saY0zI1FOlfVNdXS1AyEZPYYkPERmK7UJbBnQHXsM2hfsd0EhFoBVFUWrHP+4jmPjo2dM3VqRbN2e7NYkPt8AI1/Lx1lueqqhYF9RBBzXNWu31pbC8vDw6JiamRdeTUPYNpaWlsSIS8i8u3JiPx4G/GWP2wzaUA/gCp/CYoihKkxOO+PDH7XZpyeLD3+3iFhXhig+v8AA45xzfYNzGpqqq6tnc3NwEzydcRQGsxaO4uDguJyenQ2Vl5W2h5oXrdtkfmO7ZNmDdLSLSQuPGFUVpiwwZYj/ReysE1Fd8tOSYD3/LR5cuzv7q1eBuU+IWHz17QmQkVFU5Y/36wSOPNN1aAaqrqx/ftWvXqMWLFx9OmK02lHZBtYhsqaysvG3UqFEfhZoUrvjIAUYDNYUyRORAYFWoExRFURqbxEQ46ij49FP7iT9YwKk/rdXy0aOHzVSprLTCxJ214hYf0dFWgKxda/c7dLBxLsn+fcQbmdGjR5djXe+KUm/CdbvcDLwnIrcBHUTkeuB14KYmW5miKEoQXnsNpk+HL74IL420NYgPYwLFR1SUr7iqrLRfY2Js/Q43BxzgbN9/P4wc2XRrVZTGICzLhzFmlogcB1yEjfXoBZxmjJnblItTFEXxJzUVzj47/PnugNOW6nYpLLSVXMEKC6+46Ns3sDuvt6mcm7vvtgXXxoyBiy5q+vUqyt4SbrZLmjFmPrYzraIoSqvBv8hYdbWtGdKS8I/38NI3SOcQt8vFS+/e8NRTjb8uRWkqwv0TXC8i74vI2d76HoqiKK2B+Hgn/qGiAnburH3+vsDtcqlLfLg7/ypKayVc8dETmIUts75FRF4RkZNFRAv3KorS4mnpcR9uy4c7yyVcy4eitDbCLa++3RjzmDHmcGAIsBC4C2ihHlRFURSHlh73sbduF0VpbTTE89kFyADSgIJGXY2iKEoTsDeWj++/t/EUTemu8c908RKsjomKD6UtEJb4EJHBInKHiKwC3vYMn2qM6d9kK1MURWkkGio+Vq+GI4+EP/7Rdst9443GXxuEtnzExUFmpu9cFR9KWyBcy8c3QDfgj0APY8wVxpgfRaSFxYwriqIE0lDx8dZbTgrs1q1w+ukweTIUN3I7zVCWDwh0vaj4UNoC4YqHDGPMhcaYz4wx1SIyVETuAzY25eIURVEag4bGfHwUpDj0m282flprKMsHqPhQ2ibhBpyWi0i6iFwuIvOABcAY4PKmXJyiKEpj0BDLR0kJfPWVs3/MMc72qkZuLBEq1RZUfChtk1rFh4hEi8hkEXkX2IR1u7yFDTQ9wxjzetMvUVEUZe9oiPj48ksoK7PbgwfDhRfW/xrhEirVFmyTODcqPpS2QF2Wj63Ak8CvwMHGmMHGmDuA8iZfmaIoSiORleVs5+T4doANhdvlMmFC09UKKS2FoiK7HRUFnTr5Hve3fHTu3HjvrSj7irrExyIgBTgIOEBEOtU+XVEUpeWRlubEfZSUBPZLCYZbfBx7bNOJD/9gU/++Lf36OeXgO3eG2NjGe29F2VfUKj6MMeOAvsDHwNXY6qbvAglAdJOvTlEUpZFwd3qdN6/2uRs2wPLldjsmBsaODRQfxjTOumqL9wBrCbnpJnvs9tsb5z0VZV9TZ8CpMWadMeYOT02Po7BVTauBhSJyT1MvUFEUpTEYNcrZnj+/9rkff+xsjx1r+8MkJdm6G2CtJ7t3N866aov38HLbbTZL50/a2lNpI9SrTocx5mtjzFSgK3AZMLRJVqUoitLIuMVHXZYPf5cLWHdIXa6XGTOsQPjgg/AtI9u3O9u1NY3zd8coSmumQUXCjDF7jDGvGGOOb+wFKYqiNAX+bhdj7Ot//4PPPnOOVVXBp586+xMmONtu8eFfL+S992DKFHj8cTjhBBgzBt55p+51FRQ42/7BporSVtEKpYqitAt69XIe7gUFsG4d/Oc/MHEiHH00fPKJPTZvntPHpWtXGDLEuYa7WJnb8rFzJ0yd6vt+8+bBqafCSy/Vvi63+EhJCf9+FKU1o+JDUZR2gUig9ePhh539F16wX91WkKOP9nV3hHK7/OUvkJtrtzt29M1IcVtRgqHiQ2mPqPhQFKXd4I77eO45WLzY2f/gg0CXy9FH+54fTHz873+OcAF49ll4+WVnv65y7oWFzraKD6W9oOJDUZR2g9vy8e67vsd27IAvvoCvv3bGjjrKd04w8XH11c7Y2WfDpEnQo4czVpf4UMuH0h5R8aEoSrvBbfkIxo03OiXVBw6E7t19j/uLj/x8p2BZTAw89JDdrk8jOxUfSnuk2cSHiKSKyFsiUiwi60TkrFrmjhKRL0Vkt4hsFZHLXceyRWS2iJSIyC8icnSo6yiKorjp39/W7AjF99872/4uFwgUH95CZAD77QepqXY7I8OJFcnLg4qK0O/pFh8dO4aepyhtiea0fDyK7QmTAZwNPC4i+/tPEpE04ENsT5nOQD9shVUvrwDzPcduBGaKSC3Z8YqiKJbISBgxwnfswgshOki95vqKj0GDnO2oKN+aHd5CYl9+CSedBB9+6FxILR9Ke6RZxIeIJACTgZuNMbuNMV8D/wPODTL9SuAjY8xLxpgyY0yRMWa55zoDgFHALcaYUmPMG8Biz7UVRVHqxN/1cskltoqpm4gIGDcu8Fx3BdJt22DpUmffLT4guOvlkktsPZD77x9Qk86r4kNpj0Q10/sMACqNMStcYwuBI4PMPRhYLCLfYq0ePwCXGGPWA/sDa4wxRX7XCbCgAIjIVGAqQEZGBnPmzGnQ4nfv3t3gc1s6bfneoG3fn95bw4iP7wrsB0Dv3rvZufNnBg7szmefOb3rBwzYxYIFwcugJicfxq5d0VRVwfvvFwFJAFRXL2XOnLyaeTExQ7EGWvj448UUFuazbNlYQKioiODNN3+id+9idu06ErA+mgULviAyspGaxihKS8YY0+Qv4Ahgi9/YRcCcIHNXAAXAAUAs8BDwjefYucD3fvPvAp6raw2jR482DWX27NkNPrel05bvzZi2fX96bw2joMCY3r2NiYw05q237Nivv3rrndrXDTeEPn/wYGeeiLO9eLHvvPPPd4498YQxq1f7vsdnnxlTWOjsJyY22S3vFcDPphmeE/pqX6/msnzsBpL9xpKBoiBzS4G3jDE/AYjIbcB2EelYz+soiqIE0LEjrFgBe/ZAYqIdGzDABqN6M1eCxXt46dYNli2z28ZjpIiMtOf7z/OyeTPk5Pge375dXS5K+6W5Ak5XAFEi4v7zHA4sDTJ3EeC2O7q3lwJ9RCQpjOsoiqIEJSrKER5e/vUvyMqytTqODOYQ9uAOOvXSt69NtXXjLz7WrvU9ruJDac80i/gwxhQDbwK3i0iCiBwGTAReDDL9WWCSiIwQkWjgZuBrY0yhsTEjC4BbRCRWRCYBw4A3muM+FEVpu5x8MmzcCNOn24DTUAQTH/7BplB/8aFptkp7ojlTbf8ExAHbsOmy04wxS0XkCBHZ7Z1kjPkcuAF4zzO3H+CuCXImMAbYCdwNnG6MyUNRFKUZaKj48He75OWp5UNpvzRXzAfGmHzg1CDjXwGJfmOPA4+HuE4OMK7RF6goihIGDRUfHTr4Ht++Xfu6KO2XZhMfiqIobYGGiI+tW23TOjca86G0Z7S3i6IoSj0IJj722y9wLDbWERSVlU4jOi8qPpT2jIoPRVGUeuAvPrp3h6Sk4HPd1g9/VHwo7RkVH4qiKPUgNdWm6noJ5nLxUpf48JZYBxUfSvtCxYeiKEo9iIjwtX40VHzs2QO5uc6+ig+lPaHiQ1EUpZ40hvgAWL3a2dY6H0p7QsWHoihKPRk+3Nk+9NDQ8+oSH+7CY2r5UNoTmmqrKIpST26/3QaZjhgBw4aFnhdMfMTFQWmp3Xan36r4UNoTKj4URVHqSWYmPPBA3fOCiY+RI+HbbwPHVXwo7Ql1uyiKojQR/uIjOtrXZeNGYz6U9oSKD0VRlCbCX3z07AkZGYHz4uMDy68rSltGxYeiKEoTkZRkhYWX7GxITw+cpy4Xpb2h4kNRFKWJEPG1fvTuDWlpgfPU5aK0N1R8KIqiNCHhiA+1fCjtDRUfiqIoTUjv3s72gAEqPhQFNNVWURSlSbnySliwwIqQiRMhLy9wjooPpb2h4kNRFKUJGTECFi1y9jt3Dpyj4kNpb6jbRVEUpRmJiYH4+EqfMRUfSntDxYeiKEoz07Fjhc++ig+lvaHiQ1EUpZnxFx+aaqu0N1R8KIqiNDNq+VDaOyo+FEVRmpnkZBUfSvtGxYeiKEozo5YPpb2j4kNRFKWZUfGhtHdUfCiKojQzKj6U9o6KD0VRlGYmJUWzXZT2jYoPRVGUZsZt+YiNtS9FaU+o+FAURWlm3OJDrR5Ke0TFh6IoSjOTlVVKdrbdPuqofboURdknaGM5RVGUZiYqyvDdd/D993Dssft6NYrS/Kj4UBRF2Qd07QqnnrqvV6Eo+wZ1uyiKoiiK0qyo+FAURVEUpVlR8aEoiqIoSrPSbOJDRFJF5C0RKRaRdSJyVoh5t4pIhYjsdr36uI4bzzW8x/7TXPegKIqiKMre05wBp48C5UAGMAJ4T0QWGmOWBpn7mjHmnFquNdwYs6oJ1qgoiqIoShPTLJYPEUkAJgM3G2N2G2O+Bv4HnNsc768oiqIoSstBjDFN/yYiI4FvjDHxrrGrgSONMSf7zb0VuAKoAjYDjxhjHncdN57xCOBb4EpjTE6I950KTAXIyMgY/eqrrzZo/bt37yYxMbFB57Z02vK9Qdu+P7231ktrur/x48fPNcaM2dfrUNoWzSU+jgBeN8Z0dY1dBJxtjBnnN3cwUABsBQ4C3sAKjFc8x8cC3wPxwJ3AOGCEMaayjjXkAesaeAtpwPYGntvSacv3Bm37/vTeWi+t6f56GWPS9/UilLZFc8V87AaS/caSgSL/icaYZa7db0Xk38DpwCue4196jpWLyOXALmAQsLi2BezNH4+I/NxWlX9bvjdo2/en99Z6aev3pyh10VzZLiuAKBHp7xobDgQLNvXHALIXxxVFURRFaUE0i/gwxhQDbwK3i0iCiBwGTARe9J8rIhNFpJNYDgT+DLzjOba/iIwQkUgRSQT+BWwCljfHfSiKoiiKsvc0Z5GxPwFxwDasC2WaMWapiBwhIrtd884EVmFdMi8A/zTGPO85lgG8hnW1rAGygZOMMRU0LU818fX3JW353qBt35/eW+ulrd+fotRKswScKoqiKIqieNHy6oqiKIqiNCsqPhRFURRFaVZUfCiKoiiK0qy0a/EhIpeKyM8iUiYiz7nGDxaRT0QkX0TyROR1EenmOi4i8k8R2eF5/VNEWlS6b6h78xw7SkR+EZESEZktIr1cx2JE5L8isktEtojIlc2++AYgItki8r6I7PSs+xERifIcGyEicz33O1dERuzj5dYbETlTRJZ7miqu9hTuq/Vn2ZoQkf4iskdEprvGzvI0oSwWkbdFJHVfrrG+eP6WnvHcQ5GILBCR413H28TPTlEaQrsWH0Autkrqf/3GO2Gj0bOBXtjMm2ddx6cCp2JrlQwDTgb+2LRLrTdB701E0rBpzzcDqcDP2AwiL7cC/bH3PR64VkSOa4b17i2PYTOpumEbFx4J/ElEOmBTtadjf67PA+94xlsFInIM8E/gfCAJGAusCeNn2Zp4FPjJuyMi+wNPYvs/ZQAl2J9xayIK2ID9XewI3ATM8AjltvSzU5R6o9kugIjcCXQ3xvw+xPFRwBfGmCTP/rfAc8aYpzz7FwAXGWMObqYlh43/vXn63fzeGHOoZz8BW+Z5pDHmFxHJ9Rz/2HP8DqC/MebMfXIDYSIiy4GrjDHve/bvxVbRfQMrHLsbzy+7iKwHphpjPtxX660Pnt+3Z4wxz/iN1/qzbP6VNgwRORM4DVgG9DPGnCMifweyjTFneeb0xdbz6WyMCaiM3FoQkUXAbUBn2sDPTlEaSnu3fITLWHyrse4PLHTtL/SMtQZ81u4pALca2F9EOmEtB63x3h4EzhSReBHJAo4HPsSufZHxVdmLaB33hIhEAmOAdBFZJSIbPS6lOGr5We6b1dYfEUkGbgf83Xv+97YaKAcGNN/qGhcRycCufylt4GenKHuDio86EJFhwN+Aa1zDiUCha78QSGxpcR8h8F87nv0kzzEIvLekZljX3vIl9h/3LmAj1oz9NrXfb2sgA4jG9jc6AutSGok14bf2ewO4A2vV2eg33hburQYRiQZeAp73WDba1P0pSn1R8VELItIP+AC43BjzleuQf6O8ZGC336frlkptTf52u/b9j7VYRCQCa+V4E0jAdgzthI2TCLupYQul1PP1YWPMZmPMduB+4ARa+b15An+PBh4IcrhV35sbz+/ni1jLzaWe4TZzf4rSEFR8hMATef4pcIcxxr8HzVJssKmXcJvktQR81u7xNfcFlhpjdgKbaX33lgr0BB4xxpQZY3Zg4zxOwK59mJ9Vahgt/54A8PxMNmIbKNYMe76G/Fk22wL3jnHYoO71IrIFuBqYLCLzCLy3PkAMtkllq8Hze/cM1oI12dUKorX/7BRlr2jX4kNEokQkFogEIkUk1jOWBXyOfZg9EeTUF4ArRSRLRDKBq4Dnmm3hYRDq3oC3gCEiMtlz/G/YmAhvkNsLwE1im/vtB1xEC7s3fzzWgLXANM99pwC/w8Z2zAGqgD97Uh+9nzw/3xdrbSDPApeJSBdPXM4VwCzq/lm2dJ7CPnBHeF5PAO8BE7AuipPF9n5KwMaFvNkKg00fBwYBJxtjSl3jrf1npyh7hzGm3b6waaXG73UrcItne7f75TpPgHuAfM/rHjyZQy3lFerePMeOBn7BmvTnYLMKvOfFYNNzdwFbgSv39b2Eeb8jPPeyE5s1MAPI8BwbCcz13O88bEbBPl9zPe4tGptmWgBsAR4CYuv6Wba2l+d3drpr/yxgPVCMTZdO3ddrrOf99PL83e3x+19ydlv72elLX/V9aaqtoiiKoijNSrt2uyiKoiiK0vyo+FAURVEUpVlR8aEoiqIoSrOi4kNRFEVRlGZFxYeiKIqiKM2Kig9FURRFUZoVFR+KoiiKojQrKj4UBRCRHBHZ5qmm6R27UETmhJifLSJGRHa7XguDza3HGrzXjNqb6yiKorR0VHwoikMkcHk9z0kxxiR6XsPrnt50iEX/phVFafHoPypFcbgXuNrTG6bBiMh+IvKJiOSLyK8icobr2IkiMl9EdonIBhG51XXql56vBR5LyiEicquITHed72MdEZE5InKXiHwDlAB96nj/E0RkmYgUicgmEbl6b+5VURSlIaj4UBSHn7E9Nhr8QPa4bT4BXga6AGcCj4nIYM+UYuA8IAU4EdsM71TPsbGer15ryndhvu25wFQgCcir4/2fAf5ojEkChtC6GuwpitJGUPGhKL78DdtBNj3M+dtFpMDzuho4CcgxxjxrjKk0xswH3gB+C2CMmWOMWWyMqTbGLAJeAY7cyzU/Z4xZaoypBI6r7f2BCmCwiCQbY3YaY+bt5XsriqLUGxUfiuLCGLME267+ujBPSTPGpHhe92E7mR7kEiQFwNlAVwAROUhEZotInogUAhcDaXu57A2u7VrfH5gMnACsE5EvROSQvXxvRVGUeqNR9YoSyC3APOBfDTh3A/CFMeaYEMdfBh4BjjfG7BGRB3HER7AW08VAvGu/a5A57vNqfX9jzE/ARBGJBi4FZgA9QqxVURSlSVDLh6L4YYxZBbwG/LkBp88CBojIuSIS7XkdICKDPMeTgHyP8DgQOMt1bh5QDfRxjS0AxopITxHpCFzf0PcXkQ4icraIdDTGVAC7PO+nKIrSrKj4UJTg3A4k1DnLD2NMEXAsNtAzF9gC/BOI8Uz5E3C7iBRh40tmuM4tAe4CvvG4TA42xnyCFUKLgLlYcbE3738ukCMiu7Aun7Pre4+Koih7ixgTzNKrKIqiKIrSNKjlQ1EURVGUZkXFh6IoiqIozYqKD0VRFEVRmhUVH4qiKIqiNCsqPhRFURRFaVZUfCiKoiiK0qyo+FAURVEUpVlR8aEoiqIoSrPy/8DuaCI0Oo6VAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "\n",
    "idxmax_prediction = rfe_prediction[\"ap_val\"].idxmax()\n",
    "idxmax_error = rfe_error[\"ap_val\"].idxmax()\n",
    "\n",
    "ax.plot(-rfe_error.head(max_iter_show)[\"n_features\"], rfe_error.head(max_iter_show)[\"ap_tst\"], lw=3, color=\"orange\", label=\"RFE - Error\\nContribution\")\n",
    "ax.plot(-rfe_prediction.head(max_iter_show)[\"n_features\"], rfe_prediction.head(max_iter_show)[\"ap_tst\"], lw=3, color=\"blue\", label=\"RFE - Prediction\\nContribution\")\n",
    "ax.scatter(-rfe_prediction.loc[idxmax_prediction,\"n_features\"], rfe_prediction.loc[idxmax_prediction,\"ap_tst\"], s=200, marker=\"o\", facecolors=\"none\", edgecolors=\"red\", lw=3, zorder=100)\n",
    "ax.scatter(-rfe_error.loc[idxmax_error,\"n_features\"], rfe_error.loc[idxmax_error,\"ap_tst\"], s=200, marker=\"o\", facecolors=\"none\", edgecolors=\"red\", lw=3, zorder=100)\n",
    "ax.set_title(\"Test Set\", fontsize=12)\n",
    "ax.set_ylabel(\"Average Precision\", fontsize=12)\n",
    "ax.set_xlabel(\"N Features\", fontsize=12)\n",
    "ax.set_xticklabels([-int(x) for x in ax.get_xticks()])\n",
    "ax.grid()\n",
    "ax.legend(loc='center left', bbox_to_anchor=(1, 0.5), fontsize=12)\n",
    "ax.tick_params(axis=\"both\", which=\"major\", labelsize=12)\n",
    "\n",
    "fig.savefig(\"rfe_ap_tst_.png\", dpi=200, bbox_inches=\"tight\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "dev",
   "language": "python",
   "name": "dev"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}