main.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "8b26e953",
   "metadata": {},
   "source": [
    "# Predicting Electricity consumption\n",
    "\n",
    "## Team\n",
    "Dmitri Rozgonjuk <br>\n",
    "Eerik Sven Puudist <br>\n",
    "Triin Pohla  <br>\n",
    "Andri Hõbemägi <br>\n",
    "\n",
    "**Product Owner:** Kristjan Eljand\n",
    "\n",
    "# Introduction\n",
    "The present project was a capstone project for the Machine Learning (MTAT.03.227) course. The general aim was to predict a household's energy consumption. A major side goal of our team was to try out different modeling techniques for educational purposes.\n",
    "\n",
    "## Our Approach\n",
    "Our general approach is illustrated in Figure 1, and the steps with detailed annotation are completed throughout this notebook.\n",
    "![](images/ML_project_workflow.png \"Figure 1. General Project Workflow\")\n",
    "\n",
    "It is noteworthy that although the process seems to be relatively linear, this is to illustrate the general workflow from data ingestion to Kaggle submission. In reality, the process included several changes on-the-go, since we learned about new methods and techniques during the project. However, for the sake of cohesion, we display the **general** solution within the present notebook. This means that the model goodness metrics should reflect the **relative** goodness (not the metrics achieved throughout the project process). The original models which were used for actual submissions are split between different notebooks which are stored in the `x_old_files` directory. Although those notebooks had differences in some methods (e.g., how train-test split was done, which target variable was used, etc), the present notebook distills the optimal solutions found with those methods. To that end, we found it fruitful to limit the amount of data used for training (using `prophet`, the best MAE scores were obtained when training was done from May 1st 2022), doing the 80-20 train-validation split based on time (i.e., earlier time series were used as training, later/last timeseries as validation), computing the rolling averages for the target variable and using the 3-hour rolling average as the main target feature."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "951791c7",
   "metadata": {},
   "source": [
    "## 0. Setup\n",
    "Importing the necessary libraries and setting other settings."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "627a768a",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Importing plotly failed. Interactive plots will not work.\n"
     ]
    }
   ],
   "source": [
    "# Dataframes and vector ops\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import math\n",
    "\n",
    "# Visualization\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# ML modules\n",
    "## Preprocessing\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "## Models\n",
    "from sklearn.linear_model import LinearRegression, Lasso, Ridge, ElasticNet \n",
    "from sklearn.tree import DecisionTreeRegressor\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "import xgboost as xgb\n",
    "from prophet import Prophet # Prophet time-series forecasting\n",
    "from skforecast.ForecasterAutoreg import ForecasterAutoreg\n",
    "## Metrics\n",
    "from sklearn.metrics import mean_absolute_error\n",
    "\n",
    "# MISC\n",
    "import json # used for 'holidays' extraction\n",
    "import requests # used for 'holidays' extraction\n",
    "from copy import deepcopy\n",
    "import time\n",
    "\n",
    "# Disable warnings\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4c7f9841",
   "metadata": {},
   "source": [
    "## 1. Data Import and a Quick Glimpse\n",
    "The original data set is from Kaggle, and the below code could be uncommented in order to download that data, unzip the file, and import the data tables that invlude training, test, and submissions data file. The submissions data file is a sample submission for testing that the data are submitted to Kaggle in the appropriate format."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "45ac905b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Download data from kaggle\n",
    "#!kaggle competitions download -c predict-electricity-consumption\n",
    "\n",
    "# Unzip the data\n",
    "#!unzip predict-electricity-consumption.zip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "06f2e687",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import the data\n",
    "df_train = pd.read_csv(\"data/train.csv\")\n",
    "# Rename 'consumption' to 'y_original'\n",
    "df_train = df_train.rename(columns = {'consumption': 'y_original'})\n",
    "\n",
    "df_test = pd.read_csv(\"data/test.csv\")\n",
    "df_submission = pd.read_csv(\"data/sample_submission.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "07543af8",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dimensions of training data: (8592, 13)\n",
      "Earliest timepoint (training data): 2021-09-01 00:00:00+03:00\n",
      "Latest timepoint (training data): 2022-08-24 23:00:00+03:00\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>temp</th>\n",
       "      <th>dwpt</th>\n",
       "      <th>rhum</th>\n",
       "      <th>prcp</th>\n",
       "      <th>snow</th>\n",
       "      <th>wdir</th>\n",
       "      <th>wspd</th>\n",
       "      <th>wpgt</th>\n",
       "      <th>pres</th>\n",
       "      <th>coco</th>\n",
       "      <th>el_price</th>\n",
       "      <th>y_original</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2021-09-01 00:00:00+03:00</td>\n",
       "      <td>11.2</td>\n",
       "      <td>10.3</td>\n",
       "      <td>94.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>16.7</td>\n",
       "      <td>1012.6</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.09016</td>\n",
       "      <td>0.577</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2021-09-01 01:00:00+03:00</td>\n",
       "      <td>10.7</td>\n",
       "      <td>9.6</td>\n",
       "      <td>93.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1012.6</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.09251</td>\n",
       "      <td>0.594</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2021-09-01 02:00:00+03:00</td>\n",
       "      <td>9.9</td>\n",
       "      <td>9.0</td>\n",
       "      <td>94.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1012.2</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.08890</td>\n",
       "      <td>0.685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2021-09-01 03:00:00+03:00</td>\n",
       "      <td>10.0</td>\n",
       "      <td>8.4</td>\n",
       "      <td>90.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>330.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1011.9</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.08735</td>\n",
       "      <td>1.016</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2021-09-01 04:00:00+03:00</td>\n",
       "      <td>9.0</td>\n",
       "      <td>8.1</td>\n",
       "      <td>94.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>300.0</td>\n",
       "      <td>3.6</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1011.4</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.08688</td>\n",
       "      <td>0.677</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                        time  temp  dwpt  rhum  prcp  snow   wdir  wspd  wpgt  \\\n",
       "0  2021-09-01 00:00:00+03:00  11.2  10.3  94.0   NaN   NaN  320.0   7.2  16.7   \n",
       "1  2021-09-01 01:00:00+03:00  10.7   9.6  93.0   NaN   NaN  320.0   7.2  13.0   \n",
       "2  2021-09-01 02:00:00+03:00   9.9   9.0  94.0   NaN   NaN  320.0   7.2  13.0   \n",
       "3  2021-09-01 03:00:00+03:00  10.0   8.4  90.0   NaN   NaN  330.0   7.2  13.0   \n",
       "4  2021-09-01 04:00:00+03:00   9.0   8.1  94.0   NaN   NaN  300.0   3.6  13.0   \n",
       "\n",
       "     pres  coco  el_price  y_original  \n",
       "0  1012.6   2.0   0.09016       0.577  \n",
       "1  1012.6   2.0   0.09251       0.594  \n",
       "2  1012.2   2.0   0.08890       0.685  \n",
       "3  1011.9   1.0   0.08735       1.016  \n",
       "4  1011.4   2.0   0.08688       0.677  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(f'Dimensions of training data: {df_train.shape}')\n",
    "print(f'Earliest timepoint (training data): {min(df_train[\"time\"])}')\n",
    "print(f'Latest timepoint (training data): {max(df_train[\"time\"])}')\n",
    "df_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "12885581",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dimensions of test data: (168, 12)\n",
      "Earliest timepoint (test data): 2022-08-25 00:00:00+03:00\n",
      "Latest timepoint (test data): 2022-08-31 23:00:00+03:00\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>temp</th>\n",
       "      <th>dwpt</th>\n",
       "      <th>rhum</th>\n",
       "      <th>prcp</th>\n",
       "      <th>snow</th>\n",
       "      <th>wdir</th>\n",
       "      <th>wspd</th>\n",
       "      <th>wpgt</th>\n",
       "      <th>pres</th>\n",
       "      <th>coco</th>\n",
       "      <th>el_price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2022-08-25 00:00:00+03:00</td>\n",
       "      <td>21.1</td>\n",
       "      <td>15.2</td>\n",
       "      <td>69.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>340.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>9.3</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.25533</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2022-08-25 01:00:00+03:00</td>\n",
       "      <td>20.1</td>\n",
       "      <td>15.1</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>30.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>14.8</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.19492</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2022-08-25 02:00:00+03:00</td>\n",
       "      <td>20.1</td>\n",
       "      <td>15.1</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.18853</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2022-08-25 03:00:00+03:00</td>\n",
       "      <td>18.7</td>\n",
       "      <td>17.0</td>\n",
       "      <td>90.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>11.1</td>\n",
       "      <td>1022.4</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.19947</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2022-08-25 04:00:00+03:00</td>\n",
       "      <td>18.1</td>\n",
       "      <td>17.1</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>280.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>11.1</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.21192</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                        time  temp  dwpt  rhum  prcp  snow   wdir  wspd  wpgt  \\\n",
       "0  2022-08-25 00:00:00+03:00  21.1  15.2  69.0   0.0   NaN  340.0   9.0   9.3   \n",
       "1  2022-08-25 01:00:00+03:00  20.1  15.1  73.0   0.0   NaN   30.0   6.0  14.8   \n",
       "2  2022-08-25 02:00:00+03:00  20.1  15.1  73.0   0.0   NaN  320.0   7.0  13.0   \n",
       "3  2022-08-25 03:00:00+03:00  18.7  17.0  90.0   0.0   NaN    0.0   4.0  11.1   \n",
       "4  2022-08-25 04:00:00+03:00  18.1  17.1  94.0   0.0   NaN  280.0   7.0  11.1   \n",
       "\n",
       "     pres  coco  el_price  \n",
       "0  1022.0   2.0   0.25533  \n",
       "1  1022.0   2.0   0.19492  \n",
       "2  1022.0   2.0   0.18853  \n",
       "3  1022.4   4.0   0.19947  \n",
       "4  1022.0   3.0   0.21192  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(f'Dimensions of test data: {df_test.shape}')\n",
    "print(f'Earliest timepoint (test data): {min(df_test[\"time\"])}')\n",
    "print(f'Latest timepoint (test data): {max(df_test[\"time\"])}')\n",
    "df_test.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e19b524",
   "metadata": {},
   "source": [
    "## 2. Data Preprocessing\n",
    "In this section, we are extracting information from available data. Additionally, here is where we implement data transformations and prepare the final datasets for use in models.\n",
    "\n",
    "### 2.1. Extracting Features from `time` Column\n",
    "Before we start extracting information from the `time` variable, we first need to convert the column to a `datetime64` type in pandas. Then, we should look into the time range of each data set. The goal is to see what were the date ranges (for extracting additional features, e.g., holidays), and to see if the test data immediately follows training data in time sequence."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "90f04b8f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Transform the 'time' column to a `datetime` format \n",
    "df_train['time'] = pd.to_datetime(df_train['time'], utc=True).map(lambda x: x.tz_convert('Europe/Helsinki'))\n",
    "df_test['time'] = pd.to_datetime(df_test['time'], utc=True).map(lambda x: x.tz_convert('Europe/Helsinki'))\n",
    "df_submission['time'] = pd.to_datetime(df_submission['time'], utc=True).map(lambda x: x.tz_convert('Europe/Helsinki'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "aa2c71cc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The range of training data: (Timestamp('2021-09-01 00:00:00+0300', tz='Europe/Helsinki'), Timestamp('2022-08-24 23:00:00+0300', tz='Europe/Helsinki'))\n",
      "\n",
      "The range of test data: (Timestamp('2022-08-25 00:00:00+0300', tz='Europe/Helsinki'), Timestamp('2022-08-31 23:00:00+0300', tz='Europe/Helsinki'))\n"
     ]
    }
   ],
   "source": [
    "print(f\"The range of training data: {df_train['time'].min(), df_train['time'].max()}\")\n",
    "print()\n",
    "print(f\"The range of test data: {df_test['time'].min(), df_test['time'].max()}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5fc3b2f5",
   "metadata": {},
   "source": [
    "Ok, from this we see that, indeed, the training and test data are sequential. The training dates range from 09.01.2021 to 24.08.2022, while the range for test data is 25.08.2022 until 31.08.2022. Now, we will also download the Estonian holidays table and keep only the rows for the relevant time period. (Let's also save the holidays table in case we need to use this in the future.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "f4165e5e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# NOTE! Also possible to fetch the holidays info via 'holidays' Python module\n",
    "\n",
    "# Fetch the list of holidays from 'riigipühad.ee'\n",
    "r = requests.get('https://riigipühad.ee/?output=json')\n",
    "json_str = json.dumps(r.json())\n",
    "holidays = pd.read_json(json_str)\n",
    "holidays['time'] = pd.to_datetime(holidays['date'], utc=True)\n",
    "holidays['date'] = holidays['time'].dt.date\n",
    "\n",
    "# Replace the values for holiday kind (to cat, and more readablity)\n",
    "## 1 = 'Riigipüha' = '1_sthol' # state holiday\n",
    "## 2 = 'Rahvuspüha' = '2_nathol' # national holiday\n",
    "## 3 = 'Riiklik tähtpäev' = '3_natday' # nationally celebrated day\n",
    "## 4 = 'Lühendatud tööpäev' = '4_shortday' # shortened day before holiday\n",
    "holidays['kind_id'] = holidays['kind_id'].astype(str)\n",
    "holidays['kind_id'] = holidays['kind_id'].str.replace('1', '1_sthol')\n",
    "holidays['kind_id'] = holidays['kind_id'].str.replace('2', '2_nathol')\n",
    "holidays['kind_id'] = holidays['kind_id'].str.replace('3', '3_natday')\n",
    "holidays['kind_id'] = holidays['kind_id'].str.replace('4', '4_shortday')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "1ac9d6ee",
   "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>date</th>\n",
       "      <th>h_type</th>\n",
       "      <th>is_holiday</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2021-09-12</td>\n",
       "      <td>3_natday</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2021-09-22</td>\n",
       "      <td>3_natday</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2021-10-16</td>\n",
       "      <td>3_natday</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2021-11-02</td>\n",
       "      <td>3_natday</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2021-11-14</td>\n",
       "      <td>3_natday</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         date    h_type  is_holiday\n",
       "0  2021-09-12  3_natday           1\n",
       "1  2021-09-22  3_natday           1\n",
       "2  2021-10-16  3_natday           1\n",
       "3  2021-11-02  3_natday           1\n",
       "4  2021-11-14  3_natday           1"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "holidays.rename(columns= {'kind_id':'h_type'}, inplace = True) # rename the col for later clarity\n",
    "# holidays.to_csv('estonian_holidays.csv', index=False) # for saving to your local machine\n",
    "\n",
    "# Keep only the relevant rows\n",
    "holidays = holidays[(holidays['date'] >= df_train['time'].dt.date.min()) & \n",
    "                    (holidays['date'] <= df_test['time'].dt.date.max())].reset_index(drop = True)\n",
    "# holidays['kind_id'].value_counts() # count the unique holiday types\n",
    "\n",
    "# Create the mini df for merge\n",
    "df_holidays = holidays.loc[:, ['date', 'h_type']]\n",
    "df_holidays['is_holiday'] = np.full(len(df_holidays), 1) # 1 = is a holiday\n",
    "df_holidays.head()\n",
    "\n",
    "# might be useful for calendar data: https://towardsdatascience.com/holiday-calendars-with-pandas-9c01f1ee5fee"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3fb158b0",
   "metadata": {},
   "source": [
    "Below, we are creating a helper function that extracts information from the timestamp feature."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "b8c9fa64",
   "metadata": {},
   "outputs": [],
   "source": [
    "### 2.1. Create meaningful variables from 'time'\n",
    "def wrangle_dates(df_original):\n",
    "    \"\"\"A function that turns a pd.Series datetime column to several more granular columns\n",
    "    Args:\n",
    "        df (pandas DataFrame): input data (e.g., df_train or df_test)\n",
    "    \"\"\"\n",
    "    df = deepcopy(df_original)\n",
    "    \n",
    "    # Date variables\n",
    "    df['date'] = df['time'].dt.date # date\n",
    "    df['year'] = df['time'].dt.year # year\n",
    "    df['month'] = df['time'].dt.month # month\n",
    "    df['month_name'] = df['time'].dt.month_name() # month\n",
    "    df['dow'] = df['time'].dt.dayofweek # day-of-week\n",
    "    df['dayname'] = df['time'].dt.day_name() # dayname\n",
    "    df['weekend'] = 0 # not weekend\n",
    "\n",
    "    # update weekend column values if it's a weekend day\n",
    "    for i in range(len(df)):\n",
    "        if df['dow'][i] in [5,6]:\n",
    "            df['weekend'][i] = 1 # is weekend\n",
    "            \n",
    "    # Time variables\n",
    "    ## Only the 'hour' varies, hence no need for 'time' and 'minute'\n",
    "    df['hour'] = df['time'].dt.hour.astype(int) # hour\n",
    "    \n",
    "    # Season\n",
    "    conditions = [\n",
    "        (df['month'].isin(np.arange(3,6))), (df['month'].isin(np.arange(6,9))), \n",
    "        (df['month'].isin(np.arange(9,12))), (df['month'].isin(np.arange(1,3))), (df['month'] == 12)]\n",
    "    choices = ['Spring','Summer', 'Autumn', 'Winter', 'Winter']\n",
    "    df['season'] = np.select(conditions, choices)\n",
    "    \n",
    "    # Drop numeric day-of-week, month variables\n",
    "    df = df.drop(columns = ['dow', 'month'], axis = 1)\n",
    "    \n",
    "    # Can optionally drop the `time` variable here\n",
    "    #df = df.drop(columns = ['time'], axis = 1)\n",
    "    return df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a2018067",
   "metadata": {},
   "source": [
    "Apply the funciton `wrangle_dates()` to both training and test data and make some other preprocessing steps (impute the missing feature values,  etc)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "c642a211",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>temp</th>\n",
       "      <th>dwpt</th>\n",
       "      <th>rhum</th>\n",
       "      <th>prcp</th>\n",
       "      <th>snow</th>\n",
       "      <th>wdir</th>\n",
       "      <th>wspd</th>\n",
       "      <th>wpgt</th>\n",
       "      <th>pres</th>\n",
       "      <th>...</th>\n",
       "      <th>el_price</th>\n",
       "      <th>y_original</th>\n",
       "      <th>year</th>\n",
       "      <th>month_name</th>\n",
       "      <th>dayname</th>\n",
       "      <th>weekend</th>\n",
       "      <th>hour</th>\n",
       "      <th>season</th>\n",
       "      <th>h_type</th>\n",
       "      <th>is_holiday</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2021-09-01 00:00:00+03:00</td>\n",
       "      <td>11.2</td>\n",
       "      <td>10.3</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>16.7</td>\n",
       "      <td>1012.6</td>\n",
       "      <td>...</td>\n",
       "      <td>0.09016</td>\n",
       "      <td>0.577</td>\n",
       "      <td>2021</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2021-09-01 01:00:00+03:00</td>\n",
       "      <td>10.7</td>\n",
       "      <td>9.6</td>\n",
       "      <td>93.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1012.6</td>\n",
       "      <td>...</td>\n",
       "      <td>0.09251</td>\n",
       "      <td>0.594</td>\n",
       "      <td>2021</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2021-09-01 02:00:00+03:00</td>\n",
       "      <td>9.9</td>\n",
       "      <td>9.0</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1012.2</td>\n",
       "      <td>...</td>\n",
       "      <td>0.08890</td>\n",
       "      <td>0.685</td>\n",
       "      <td>2021</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2021-09-01 03:00:00+03:00</td>\n",
       "      <td>10.0</td>\n",
       "      <td>8.4</td>\n",
       "      <td>90.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>330.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1011.9</td>\n",
       "      <td>...</td>\n",
       "      <td>0.08735</td>\n",
       "      <td>1.016</td>\n",
       "      <td>2021</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2021-09-01 04:00:00+03:00</td>\n",
       "      <td>9.0</td>\n",
       "      <td>8.1</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>300.0</td>\n",
       "      <td>3.6</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1011.4</td>\n",
       "      <td>...</td>\n",
       "      <td>0.08688</td>\n",
       "      <td>0.677</td>\n",
       "      <td>2021</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                       time  temp  dwpt  rhum  prcp  snow   wdir  wspd  wpgt  \\\n",
       "0 2021-09-01 00:00:00+03:00  11.2  10.3  94.0   0.0   0.0  320.0   7.2  16.7   \n",
       "1 2021-09-01 01:00:00+03:00  10.7   9.6  93.0   0.0   0.0  320.0   7.2  13.0   \n",
       "2 2021-09-01 02:00:00+03:00   9.9   9.0  94.0   0.0   0.0  320.0   7.2  13.0   \n",
       "3 2021-09-01 03:00:00+03:00  10.0   8.4  90.0   0.0   0.0  330.0   7.2  13.0   \n",
       "4 2021-09-01 04:00:00+03:00   9.0   8.1  94.0   0.0   0.0  300.0   3.6  13.0   \n",
       "\n",
       "     pres  ... el_price  y_original  year  month_name    dayname weekend  \\\n",
       "0  1012.6  ...  0.09016       0.577  2021   September  Wednesday       0   \n",
       "1  1012.6  ...  0.09251       0.594  2021   September  Wednesday       0   \n",
       "2  1012.2  ...  0.08890       0.685  2021   September  Wednesday       0   \n",
       "3  1011.9  ...  0.08735       1.016  2021   September  Wednesday       0   \n",
       "4  1011.4  ...  0.08688       0.677  2021   September  Wednesday       0   \n",
       "\n",
       "   hour  season   h_type is_holiday  \n",
       "0     0  Autumn  not_hol          0  \n",
       "1     1  Autumn  not_hol          0  \n",
       "2     2  Autumn  not_hol          0  \n",
       "3     3  Autumn  not_hol          0  \n",
       "4     4  Autumn  not_hol          0  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_train = wrangle_dates(df_train)\n",
    "\n",
    "# Add holidays info\n",
    "df_train = df_train.merge(df_holidays, on = 'date', how = 'left')\n",
    "df_train['h_type'] = df_train['h_type'].fillna('not_hol').astype(str)\n",
    "df_train['is_holiday'] = df_train['is_holiday'].fillna(0).astype(int)\n",
    "\n",
    "# Custom fill for `snow` (fill with the next non-NaN value)\n",
    "snow = df_train['snow'].to_numpy()\n",
    "snow_index = df_train.index[~df_train['snow'].isna()].tolist()\n",
    "current_snow_value = 0\n",
    "for i in range(len(snow)):\n",
    "    if i > 5241:\n",
    "        current_snow_value = 0\n",
    "        snow[i] = current_snow_value\n",
    "    if i not in (snow_index):\n",
    "        snow[i] = current_snow_value\n",
    "    if i in (snow_index):\n",
    "        current_snow_value = snow[i]\n",
    "        snow[i] = current_snow_value\n",
    "df_train['snow'] = snow\n",
    "\n",
    "# Fill `prcp` with 0\n",
    "df_train['prcp'] = df_train['prcp'].fillna(0.0)\n",
    "\n",
    "# Fill NAs with last valid values, convert `coco` to categorical\n",
    "df_train['coco'].fillna(method = 'ffill', inplace = True)\n",
    "df_train['coco'] = df_train['coco'].astype(str)\n",
    "\n",
    "# Remove 'date', as it's redundant\n",
    "df_train = df_train.drop(columns = ['date'])\n",
    "\n",
    "df_train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d9db0674",
   "metadata": {},
   "source": [
    "The same preprocessing for test data set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "7a8f4939",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>temp</th>\n",
       "      <th>dwpt</th>\n",
       "      <th>rhum</th>\n",
       "      <th>prcp</th>\n",
       "      <th>snow</th>\n",
       "      <th>wdir</th>\n",
       "      <th>wspd</th>\n",
       "      <th>wpgt</th>\n",
       "      <th>pres</th>\n",
       "      <th>coco</th>\n",
       "      <th>el_price</th>\n",
       "      <th>year</th>\n",
       "      <th>month_name</th>\n",
       "      <th>dayname</th>\n",
       "      <th>weekend</th>\n",
       "      <th>hour</th>\n",
       "      <th>season</th>\n",
       "      <th>h_type</th>\n",
       "      <th>is_holiday</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2022-08-25 00:00:00+03:00</td>\n",
       "      <td>21.1</td>\n",
       "      <td>15.2</td>\n",
       "      <td>69.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>340.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>9.3</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.25533</td>\n",
       "      <td>2022</td>\n",
       "      <td>August</td>\n",
       "      <td>Thursday</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Summer</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2022-08-25 01:00:00+03:00</td>\n",
       "      <td>20.1</td>\n",
       "      <td>15.1</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>14.8</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.19492</td>\n",
       "      <td>2022</td>\n",
       "      <td>August</td>\n",
       "      <td>Thursday</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>Summer</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2022-08-25 02:00:00+03:00</td>\n",
       "      <td>20.1</td>\n",
       "      <td>15.1</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.18853</td>\n",
       "      <td>2022</td>\n",
       "      <td>August</td>\n",
       "      <td>Thursday</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>Summer</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2022-08-25 03:00:00+03:00</td>\n",
       "      <td>18.7</td>\n",
       "      <td>17.0</td>\n",
       "      <td>90.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>11.1</td>\n",
       "      <td>1022.4</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.19947</td>\n",
       "      <td>2022</td>\n",
       "      <td>August</td>\n",
       "      <td>Thursday</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Summer</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2022-08-25 04:00:00+03:00</td>\n",
       "      <td>18.1</td>\n",
       "      <td>17.1</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>280.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>11.1</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.21192</td>\n",
       "      <td>2022</td>\n",
       "      <td>August</td>\n",
       "      <td>Thursday</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>Summer</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                       time  temp  dwpt  rhum  prcp  snow   wdir  wspd  wpgt  \\\n",
       "0 2022-08-25 00:00:00+03:00  21.1  15.2  69.0   0.0   0.0  340.0   9.0   9.3   \n",
       "1 2022-08-25 01:00:00+03:00  20.1  15.1  73.0   0.0   0.0   30.0   6.0  14.8   \n",
       "2 2022-08-25 02:00:00+03:00  20.1  15.1  73.0   0.0   0.0  320.0   7.0  13.0   \n",
       "3 2022-08-25 03:00:00+03:00  18.7  17.0  90.0   0.0   0.0    0.0   4.0  11.1   \n",
       "4 2022-08-25 04:00:00+03:00  18.1  17.1  94.0   0.0   0.0  280.0   7.0  11.1   \n",
       "\n",
       "     pres coco  el_price  year month_name   dayname  weekend  hour  season  \\\n",
       "0  1022.0  2.0   0.25533  2022     August  Thursday        0     0  Summer   \n",
       "1  1022.0  2.0   0.19492  2022     August  Thursday        0     1  Summer   \n",
       "2  1022.0  2.0   0.18853  2022     August  Thursday        0     2  Summer   \n",
       "3  1022.4  4.0   0.19947  2022     August  Thursday        0     3  Summer   \n",
       "4  1022.0  3.0   0.21192  2022     August  Thursday        0     4  Summer   \n",
       "\n",
       "    h_type  is_holiday  \n",
       "0  not_hol           0  \n",
       "1  not_hol           0  \n",
       "2  not_hol           0  \n",
       "3  not_hol           0  \n",
       "4  not_hol           0  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test = wrangle_dates(df_test)\n",
    "\n",
    "# Add holidays info\n",
    "df_test = df_test.merge(df_holidays, on = 'date', how = 'left')\n",
    "df_test['h_type'] = df_test['h_type'].fillna('not_hol').astype(str)\n",
    "df_test['is_holiday'] = df_test['is_holiday'].fillna(0).astype(int)\n",
    "\n",
    "# Fill `snow` with 0 (no snow in August)\n",
    "df_test['snow'] = df_test['snow'].fillna(0.0)\n",
    "\n",
    "# Fill `prcp` with 0\n",
    "df_test['prcp'] = df_test['prcp'].fillna(0.0)\n",
    "df_test.head()\n",
    "\n",
    "# Fill NAs with last valid values, convert `coco` to categorical\n",
    "df_test['coco'].fillna(method = 'ffill', inplace = True)\n",
    "df_test['coco'] = df_test['coco'].astype(str)\n",
    "\n",
    "# Remove 'date', as it's redundant\n",
    "df_test = df_test.drop(columns = ['date'])\n",
    "\n",
    "df_test.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4a891249",
   "metadata": {},
   "source": [
    "## 2.1. `consumption` transformations\n",
    "This is our target variable. Here, we first impute the few missing data points by using interpolation - in essence, this is the value of average of the previous and next time point. In order to avoid a zero when doing the log-normalization later, we also add a tiny constant to all values (to avoid zeros)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "97e47256",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fill the NAs with the average of previous and next value in time-series\n",
    "df_train['y_imp'] = df_train['y_original'].interpolate()\n",
    "\n",
    "# Add a constant to avoid '0.0'\n",
    "df_train['y_imp'] = df_train['y_imp'] + 0.0000000000001"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7a82412c",
   "metadata": {},
   "source": [
    "We should also try to smooth our data a bit - using moving averages could be useful here. To that end, let's explore some potential autocorrelation values of target feature and, based on that, adjust the values with a N-day moving average. \n",
    "\n",
    "First, let's compute a vector of autocorrelations. This should help us with selecting how many shifts (i.e., the N hours) would be reasonable for predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "6c6fa298",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 500x200 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "autocors = [df_train['y_imp'].autocorr(i) for i in range(10)]\n",
    "\n",
    "plt.figure(figsize=[5,2])\n",
    "plt.plot(autocors)\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0e6d94c0",
   "metadata": {},
   "source": [
    "The figure above shows that perhaps it makes sense to compute the rolling averages up to three hours the most, as after that the autocorrelations get quite low."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "f64843f6",
   "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>time</th>\n",
       "      <th>temp</th>\n",
       "      <th>dwpt</th>\n",
       "      <th>rhum</th>\n",
       "      <th>prcp</th>\n",
       "      <th>snow</th>\n",
       "      <th>wdir</th>\n",
       "      <th>wspd</th>\n",
       "      <th>wpgt</th>\n",
       "      <th>pres</th>\n",
       "      <th>...</th>\n",
       "      <th>month_name</th>\n",
       "      <th>dayname</th>\n",
       "      <th>weekend</th>\n",
       "      <th>hour</th>\n",
       "      <th>season</th>\n",
       "      <th>h_type</th>\n",
       "      <th>is_holiday</th>\n",
       "      <th>y_imp</th>\n",
       "      <th>y_imp_r2</th>\n",
       "      <th>y_imp_r3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2021-09-01 00:00:00+03:00</td>\n",
       "      <td>11.2</td>\n",
       "      <td>10.3</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>16.7</td>\n",
       "      <td>1012.6</td>\n",
       "      <td>...</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>0.577</td>\n",
       "      <td>0.5770</td>\n",
       "      <td>0.577000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2021-09-01 01:00:00+03:00</td>\n",
       "      <td>10.7</td>\n",
       "      <td>9.6</td>\n",
       "      <td>93.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1012.6</td>\n",
       "      <td>...</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>0.594</td>\n",
       "      <td>0.5855</td>\n",
       "      <td>0.594000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2021-09-01 02:00:00+03:00</td>\n",
       "      <td>9.9</td>\n",
       "      <td>9.0</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1012.2</td>\n",
       "      <td>...</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>0.685</td>\n",
       "      <td>0.6395</td>\n",
       "      <td>0.618667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2021-09-01 03:00:00+03:00</td>\n",
       "      <td>10.0</td>\n",
       "      <td>8.4</td>\n",
       "      <td>90.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>330.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1011.9</td>\n",
       "      <td>...</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>1.016</td>\n",
       "      <td>0.8505</td>\n",
       "      <td>0.765000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2021-09-01 04:00:00+03:00</td>\n",
       "      <td>9.0</td>\n",
       "      <td>8.1</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>300.0</td>\n",
       "      <td>3.6</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1011.4</td>\n",
       "      <td>...</td>\n",
       "      <td>September</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>0.677</td>\n",
       "      <td>0.8465</td>\n",
       "      <td>0.792667</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 24 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                       time  temp  dwpt  rhum  prcp  snow   wdir  wspd  wpgt  \\\n",
       "0 2021-09-01 00:00:00+03:00  11.2  10.3  94.0   0.0   0.0  320.0   7.2  16.7   \n",
       "1 2021-09-01 01:00:00+03:00  10.7   9.6  93.0   0.0   0.0  320.0   7.2  13.0   \n",
       "2 2021-09-01 02:00:00+03:00   9.9   9.0  94.0   0.0   0.0  320.0   7.2  13.0   \n",
       "3 2021-09-01 03:00:00+03:00  10.0   8.4  90.0   0.0   0.0  330.0   7.2  13.0   \n",
       "4 2021-09-01 04:00:00+03:00   9.0   8.1  94.0   0.0   0.0  300.0   3.6  13.0   \n",
       "\n",
       "     pres  ... month_name    dayname  weekend  hour  season   h_type  \\\n",
       "0  1012.6  ...  September  Wednesday        0     0  Autumn  not_hol   \n",
       "1  1012.6  ...  September  Wednesday        0     1  Autumn  not_hol   \n",
       "2  1012.2  ...  September  Wednesday        0     2  Autumn  not_hol   \n",
       "3  1011.9  ...  September  Wednesday        0     3  Autumn  not_hol   \n",
       "4  1011.4  ...  September  Wednesday        0     4  Autumn  not_hol   \n",
       "\n",
       "   is_holiday  y_imp y_imp_r2  y_imp_r3  \n",
       "0           0  0.577   0.5770  0.577000  \n",
       "1           0  0.594   0.5855  0.594000  \n",
       "2           0  0.685   0.6395  0.618667  \n",
       "3           0  1.016   0.8505  0.765000  \n",
       "4           0  0.677   0.8465  0.792667  \n",
       "\n",
       "[5 rows x 24 columns]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Compute new values: rolling means over 2 and 3 days\n",
    "## Fill the NAs of N rows (N = size of rolling window) with the original values\n",
    "df_train['y_imp_r2'] = df_train['y_imp'].rolling(2).mean()\n",
    "nan_idx = df_train[df_train['y_imp_r2'].isnull()].index\n",
    "df_train.loc[nan_idx.values, 'y_imp_r2'] = df_train.loc[nan_idx.values, 'y_imp_r2'].fillna(df_train.loc[nan_idx.values, 'y_imp'])\n",
    "\n",
    "df_train['y_imp_r3'] = df_train['y_imp'].rolling(3).mean()\n",
    "nan_idx = df_train[df_train['y_imp_r3'].isnull()].index\n",
    "df_train.loc[nan_idx.values, 'y_imp_r3'] = df_train.loc[nan_idx.values, 'y_imp_r3'].fillna(df_train.loc[nan_idx.values, 'y_imp'])\n",
    "\n",
    "df_train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5f2efb61",
   "metadata": {},
   "source": [
    "Let's take a look at the different rolling means values:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "91c52f9d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0IAAAGsCAYAAADwjxevAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8qNh9FAAAACXBIWXMAAA9hAAAPYQGoP6dpAAEAAElEQVR4nOydd3gUVffHv5uEhASS0ENHREQFVIpSBBFREPH92Qu89vLq+9rR11dsYAUVxYqIdBBQpCodpJdAQgIJkARSIL0nm7qb7N7fH2E3s7szOzO7M7uzu+fzPHme7LR75t6ZO+fcc+65OsYYA0EQBEEQBEEQRAAR5G0BCIIgCIIgCIIgPA0ZQgRBEARBEARBBBxkCBEEQRAEQRAEEXCQIUQQBEEQBEEQRMBBhhBBEARBEARBEAEHGUIEQRAEQRAEQQQcZAgRBEEQBEEQBBFwhHhbACmYzWbk5eUhMjISOp3O2+IQBEEQBEEQBOElGGOoqqpC165dERTkul/HJwyhvLw89OjRw9tiEARBEARBEAShEbKzs9G9e3eXz/cJQygyMhJA081GRUV5WRqCIAiCIAiCILyFXq9Hjx49rDaCq/iEIWQJh4uKiiJDiCAIgiAIgiAIt6fMULIEgiAIgiAIgiACDjKECMIHaDSZ8d3uc4jLKvO2KARBqMCBc8XIKK72thgEQRABhduG0P79+/GPf/wDXbt2hU6nw4YNG2z2M8YwY8YMdO3aFeHh4bjllltw+vRpd4sliIDit7hsfL0zDQ/MO+JtUQiCUJhTORV4bOEx3PrVPm+LQhAEEVC4bQjV1NTguuuuww8//MC7/4svvsDXX3+NH374AcePH0fnzp1x++23o6qqyt2iCUIRymuM+HJ7CtI1PBqbXlTjbREIglCJ5Fy9t0XQJNlltag1NnpbDIIg/Bi3DaGJEyfik08+wX333eewjzGGb775Bu+++y7uu+8+DBgwAEuXLkVtbS1WrlzpbtEEoQjvrE/Cj3vSMfHbA94WRRBaPovwBnVGE/anFcPQaPK2KH5NEL3fDpwrrMLoL/bg5i/2eFsUgiD8GFXnCGVmZqKgoADjx4+3bgsLC8OYMWNw+PBhwfMMBgP0er3NH0GoxYmL5QAAY6PZy5IQhLZ4/bdEPL7oGD7+64y3RfFrgsgScmDX2SIAQEm10cuSEAThz6hqCBUUFAAAYmJibLbHxMRY9/Exc+ZMREdHW/9oMVXCHsaYt0VQnF1nCjFj02k0mBwNMlKTCG+w7XRTP73i6EUvS+Lf0PtNEAThHTySNc4+xzdjzGne72nTpqGystL6l52drbaIhA/RYDLjzu8O4qWVJ7wtiqI8uywOSw5nYfUxUjoJIpAIothXgiAIr6CqIdS5c2cAcPD+FBUVOXiJuISFhVkXT6VFVAl7jmeW4Wy+Hn+dylfkejqNjcfmV9Z7WwSCIDxIEC1k4QDZhgRBeAJVu9/evXujc+fO2Llzp3Wb0WjEvn37MHLkSDWLJvwYTwfFpRdX49mlcTiZXeGR8vjuz5lSUFRVj+TcStXkIaSRXVaLz7acRV5FnbdFIXjILKnB+xuSkVNe621RHCCPEEEQhHcIcfcC1dXVOH/+vPV3ZmYmEhMT0a5dO/Ts2ROvvfYaPvvsM/Tt2xd9+/bFZ599hoiICEyZMsXdognCIzy7NA6ZJTXYdbYQWbMm2ewzm5niE535pj85CyW98dPdAIBtr43GVZ3Je6oERVX16Ng6zGm92/PEomPIKKnB/rRibHvtZhWlI1zh4Z+PoKjKgNjMUux4fYy3xbFBznNGEARBKIfbHqG4uDgMGjQIgwYNAgBMnToVgwYNwgcffAAAeOutt/Daa6/hP//5D4YOHYrc3Fzs2LEDkZGR7hZNEB7hQin/Gj7f7jqHwZ/sFNzvaeIvlHtbBL9gW3I+bvx0N97645Ss8zJKmp6DlALvrZFWUm3A22tPSfZeltcYccc3+/HzvnR1BdMARVUGAEBaoe16YfUNJjyz5Dh+jb3gDbEAUPpsX8TQaEJ2mfa8i0qQVVKD+fvTUWOgNZwI/8dtQ+iWW24BY8zhb8mSJQCaRrpmzJiB/Px81NfXY9++fRgwYIC7xRKExxAarZ2zKw0VtQ34YnuqouUxjwf/EVzm7DwHAFgTn+NlSeTzzrokrD6ejbt/PCTp+Hn70pFSUIWZW1NUlky7rDh6AbtTivDu+mSvyeDNeYr1DcJrRJ0rrMLrvyUiQ8OLTXuL++Yexugv9uB4Vpm3RVGc2+fsw2dbUvD5Nv/vFw6nl+DWr/biSHqpt0UhvARN0SQId/GA3UIDxoQUzhfJU1gNKq+dtTExF6+tTtD0gqz6eu+PensrMm7xoUxc9f42bEniTzzz4M9HsD4hF48tPOZhyTzPvrRiTJizH6dyKiQdfzqvaX3DdSd8b8BEjAZT00ftWKb/GXn2TPklFhnFNZj8y1Fvi0J4CTKECL8kp7wW+voGSceKKSFiOoriHhzebAnKFuFvHE4v8dtR6x/3nMdHf0pc0FRjz8mrqxOxITEPq2IpJbwW+fDSc/Xq6gTe/RW1TX1obgAkAHli0TGkFlbh8UX+b/QRBNGM28kSCEJr5JTXYtTne6DTAZkzJ4mfQPg0qQVVmPJLLAA4JLNIL65GRGgwukSHS76e1uatf3kp9HLyjT3QN8Y351aW1Ri9LQLhBC2uT+2t17BapodQi3WnFP58bwRhgTxChN8Rl9WUNECpTtzTijG/Q0hj2rlMcivqsCYuGw0m5UOxzubrebeX1Rgx7qt9GDHzb8XLVBqTWfxhrXMyl0PrkD7lHG+/3dQ+BEEEKmQI+TB/xOfg213nPFomY8zv1klxVwlRetSM+eEw3Piv9+G/f5zC/P0Zil9byFDVSjY/MX7am46rP9iGpBzna0H54WNBED6PP7+XlLiHCATIEPJh3lxzEnN2pXl0Mc2ZW1MwctbfWHwo02Nl2uPpD49cb0yDyYxTORUwSxjl58MfP6w1xiZvxt7UIi9Loj0+35YCY6MZH2xyP2uZtz0LhG/ij4MvBEEQUiBDyA+wTGj1BJYR/U83n/VYmXJRPJRN5Hr2OsRbf5zC//1wCN/9rYy3LqO4GjvOFChyLW9jVDlLmZbJLqtFosT1ffwNqXr22Xw9DpwrVlcYP8bYaMYbv5/EhoRcWedp0QzS2lw9IdT2mhzLLBMM/1Ubso+JQIAMIS9SVd+AI+mlLnsOLHjDfU39ozDrLykhP+4579L59nV761f7kFHsG2FeYhhNnntyhNZ/Uus8MUZ/sQf3/HgIWSWutaUvv3NS+6iJ3x7AYwuP+W0GQCGUeuT+iM/B2hM5eO23RFnnkcKrTQoq6/HQz0cw8dsD3haFIPwWMoS8yIPzjmDyL0fdXtFcrY9YZW2DYMhEIIVSuJo+u8FFpd+fq9by3Bw8V4L75h5CWmGV29cUMly0+oymFHhndNeXyHTRWPRdlLGEymoMilyHkI6a3UxepX/NxyUILUKGkBdJKWhSAtfLDGPwBCculuO6j3bgP7+e8LYoDvhKyISrBMIE1UcXxuLExQo8tyzO7Wv5+eNgRYphp5Y3y13kKosatWEJETYm5uLWr/binAIDHN5CS49ei6BmFc3dyBGCIPghQ0gDuNu9qdE9LjjQNBdoazL/3BR/6pLFlEcx3dJVpa3BZMb5InVDgGoMjZoeXS+rVm99Ga0aBVrCU1XkT/2FlvH2M//q6kRkFNdg6u8nvSqHJ1Hz2Q4Jbm7PBrPn51fSe0sEAmQIEYSbuLo2zvPL43Hb1/vwR3yOwhI1M3b2XoydvVc0NbMr/HUqD3f/eAiF+nrFr+1NfNl88lXZN53Mw1TOvBZSwHwboTWvtiUX4FhmmaRr+PraaUrQgmMINXpwfiVBBBJkCKlEnVH64ofuhoF4Yy6ElkNXPD0quifVtSxXf6c0pZJeeNA2FbmSdVtU1TRnYOfZQuUuiqZn7qWVCTiZXSErfFKN58ZfHD9idaPhV85tXlmVgHWcEGGtzu9SC+4zfCqnwi+zK14srcULK+Lx0M9HvC2Koqj5qHK/Zd54IwLtPSQCEzKEVOD349m4+oNtWHXsoqTj1ehqGGN4dmkcpv6eqMLVCSXxyMdG4TIMHEUtyMuGiJSRYzl1rLZh5WpTqPGYeErPkT1HyMk+k5nhqx2pOHS+xC2ZtMr//XAIt3y5x9tiKE6+j0z81+q4ChklBKEOZAipwFtrTwEApq1L8kh5fN1jRkkNdp0txLoTuTD52CRLs5lhwYEMxF+QFkKhNr4SomFmQKOLYXpy4X6Tr+gU6ZEy/R0+A0yu8qNV75jcBCDObnttfA6+//s8/rkg1k2ptEtepX+FmwLen7+kFmomt/F2jfmW5kAQrkGGkBZwOzbOcVO9QIy2L/DnqTx8svks7v/Jv0IopOLqyN/7G5Ix6vM9vHOWlP6gaSmznRT9yhcGU70ho3Z1U+HKyCrVbvIPV9FsM7gIXx+m3WfNN/CBLowgfBIyhDSA+1njHK/gyzHmJ7Pdm9iv9PfW0x9wd56HAn090j2wGCXXySinftQwoKQUL6dU7SpsEtJna1WlJi0u4PF2CK09hkYTlh+9gAvuGtb+/Gz7870RxCVCvC0AoQ5mzohc0+icxr5CTtBaLLmna843vBfaF5Lbblp6B1ytOR+ocsVwdq/aNVS9j7brRp5wat/Lz/sy8PXONABA1qxJ6hamAIH0/hOEJyGPkAZwP2ucyH73Lu9V+BRuf/sg2N+PGl4TpeuMezmt6l6u3rKWvCr+9qxLJUBv2wZfGGyQA9ew0cK9xWaW8m6XK5nH7sQLVeb9ViII9SFDKABw5Zvjze+UBr6RNqg9yVdL822koqXMZ5LmCClfrMto7fn2BLKVywCrI74+xteS3HDhk5x7hz58awRB+BlkCGkAdxVhf1YaXLk3bYeHqA+fR0NxY0tTzxx/g9uGxnlGEjWRcgtaffblegB8cXBAafytBrjGntkfXshLqOndsl1HyBvrBfpPOxGEEGQIaQAVksbZ7ZdfgFYVKjUQu1e1q8IhNM4Hvj2KfZQD6DlzBR94FCQhex0hf7lxiQTCa8BNluBPhpCaePu5oFYiAgEyhPwA/lEbzkiSD/dmPiy6y/jCPbscGie6QR3kGG5aHQTw5fdYLgF0q4K45A33uup8CR7ZdX7yTfIWVGcEoQ5kCGkAdTxCvttrylFElVgvKcjLLiFfbCktyewv6wiJQWEqTQRKNfhbeKBOYx4hpUTw/p0QBOEOZAhpALfXEVKhJ9bAdwqA+MJ8SbmOaw7JHRUN8fICF/b36AvtyZXZLQ+KAlWvkTFwychRcP3F+JGfLME/7lsqWvVCqkWANa8iUJURhDqQIeQHvLAiHm+vPWW31XfDEDwtb5CIIeR5HUWF9Nkavx5BSMUfjQa+Ps+fE8UI3dqPe87jP7/GaypjXkm1AStjL6La0Mi7X83vlbfb09d0B4JwBTKENIASo5+rj2crIEkz3uyAbdab8EB5XvcI2f/2gY8PV0Y58qox0i8lvbmcYrWqS0rLGqdN6X3hmfYmVD1NfLk9FVuSCrA/rdjbolh5dEEs3lmfhPc3JHtVjkDzkhKEpyBDKABwJdZcK33ukXT+Re+URGyOkJhyeXnHVu4JoHBde0IX5j5T3n5UhG7X1qD2tpTi8EmofanVQSv9jzfx5zoQU+qVmPupFCkFVQCAbckFvPv9uJl8ot8kCHchQygA8OUPalmNUfY5cg2BYLHQOLHrKT7/RtnrqYKLHiFPolW5AhG5CpWzyfT+2K7a9OO5jviSDuKovpC1zOfI5OUHzw8fe4LQBGQIaQB//LArBZ9CxN2kxKfS3e+t28ku3DxfUhkaWU/VW4+6lt4xIVlE7W0J9+BvCnWgwO8N1NBDSwh6sTwVsuaNPkxL/SZBqAUZQhpA7Q+er/Vl13ZvY/3flTmzcpVBbyuPDlnjVEmWoOw1uSJfLKvB9tMFXothFzJkXZ5r5qV5NrzKsK+9vAIouaCqRqdBuQXfu+PaOkK+gRaea/leSpUEcYIW6okg/B0yhDSAK52dHKXTFQXVm8pGr/YR1v/Nnvj6uHmzgTiJlatEHDpfiueXx+PAuRKvyKL0s6r2o++6N038TH8xEpzdqT++bv52S0r0iVp7lIUy2Xmq7bzhIfTHd40g7CFDSANosa/RSgfIFxqn9GJ8Yh9ctacI+XrWOAun8/QSTlReFin4grEqqvipcAty19wiPIcrze0zhrAGXkd69gmCAMgQ0gSuKGmyUhbLvrp24BuE424S+/BLqVutKQ++0F58MgZroDcRjOP3sBzOkOXN1ZTknsNZHWntfVUC/nWE/LfttfBcKyaD51xCBEGogAZUF0JtfPl7yp8sgZO6mefebOaGKDDBXCx7keKJCNy8Hq+0isuonYdKJ2HxYA2JG5DIfV4Cr7kC744JceipIAj1IUNIA7jS2ck6x4d7U95JxHLOl3CMu2la3R1Z9EUlnX8E2/NyALCx/JQQwVseB2VSDishiQbwwXfCHfjWMvPlKhB9ljVwc0rJ4CnvljeqTEsDXgShFmQIEbx4U6HiFs07P1XhvtndOUJKo07WuMBA8MOtoQpwVbcIJJ3E2Tvgj/UQFd7CYZs/3qdU/MagJwhC85AhpAVUyBpnmzpY3TlInkZOsgRfmCPk0D4arnsLLivzyooBwNZQtZk/xg2Zk1Gytx4HvnK1/B7KQe5t+Mt9S4X3fl1Kn+0bFoQveIykoqasYmHgBEG4DxlCGkDt/s3fOlCbBVV5v/tcBVgcMeVBPCGDhEJUPN8eTxh2fIaF1yLjBG5YCxOy+RCSS5vSag9vD1wQ7kMhV/LRan9GEL6O6obQjBkzoNPpbP46d+6sdrF+TyB3ibLmCEmzhNw6QGvfdE9koHI9vEvdyvKFZAmue9MkeDc16hGQvaCqgtfyVXxZ8VWijbxl8AqtFySEPz+PfnxrBGElxBOF9O/fH7t27bL+Dg4O9kSxPoPqyqGqV1cX/hXXlb0jb6uO9rfjC+2llIxK1z1XedTJ9Ax6Cn9WnJSC6si1OvAVb5kWmlcLMsiB3gmCUAePGEIhISHkBXKCS1njZJzkb2EI7VqFWv9vHeY4yZiLpFF0EeXB08qFu+3FJ6/Sj4BSi9oqkuWNez1Bj5CMOUIqNLhNrL/L1xA/RquKsC97NzxBoNVPdlktOrQO87YYiqBm23n7qfAz1YEgePHIHKFz586ha9eu6N27Nx555BFkZGQ4Pd5gMECv19v8EYGDmDIXFNR8QBDPsfLXEXIzfbbSYWfunu+Bj5emkiVoVPnnohWFotFkxrbkApRUG7wtilMCzTDgw5drQKz99qQUiV7DB15rj+LLzwNBaBnVDaFhw4Zh2bJl2L59O3755RcUFBRg5MiRKC0tFTxn5syZiI6Otv716NFDbTG9iitKktiHhntNf+5APbFujNofZK0ZUq6W4lIojwKSCC2oyn1H5IjmqkxSDTI57a20AbXoUCZeWBGPu384pOyFFcbZNA1fMHyVwN88+VzE7syX2tiPm4kGJIiAQHVDaOLEibj//vsxcOBA3Hbbbdi8eTMAYOnSpYLnTJs2DZWVlda/7OxstcX0Kmp3Nkp21B4PE+PbJnI/UkKlbI53857c9uDY/1bhcVD6kr728fe2vEzgf1evIYTYo7w1uQAAkFtR56IUrqFk/Xu7LQnXsMn2KeNYogl/NowJwpt4ZI4Ql1atWmHgwIE4d+6c4DFhYWEIC/OP+GG1EO8TA6PTVCQ7kZfTZztcz93zPREap34RLmHjBdKukK6dpsANabZO7PEZQRWCL9Oj56UILBSqYDXbyduvgbfLJwhP4PF1hAwGA86ePYsuXbp4umjNonZno6THyTMOIdHxQs5/IiGCCiRLUBvlDSllwtaclqEhJULKnDBZC6q6+Dw4O802WYKM0DiF1SylklzIRW6ppH/5nxIqJ1TV232yFvG354EgtILqhtCbb76Jffv2ITMzE7GxsXjggQeg1+vxxBNPqF20z6BKKJQSsTg8qJFRS0084THSGp7xCLlWiBqy2YRCCvyvJc1a1ZXonez75K8zOJVTqV7hhNfRcv/sr4q8v94XoKlukyBUQ/XQuJycHEyePBklJSXo2LEjhg8fjqNHj6JXr15qFx3QqNWBaeEzazMhnudGucqApHkVbioP7o7a25/vC7HgSomo+DpCXM+LwvWYU16LjOIa3HxlR9nnKiGJu9dYcDBTASlcQ/aCqtp/BVTHlyequ9t+jGnbqCMIwn9Q3RBavXq12kUEJLLWEVJPDK8jVg9SlGGxz63ac4SUXlCVTx6llSpPlCEZCfqSEpKN+nwPAODXZ4fhpis6uHwdObKoaRCcK6zSbAiSLwwGKAnv3fpZFfjZ7XgI7855DLDXkAhQPD5HiHBEjY++mNfEVTy/uCjPNoXL0Jwy6K5hpZLKUVFrxK+xF1Cor0dRVb0qZbiLUGics4VWD6eXoLzGaN0mFgoZl1XOu93ZCLak91HUqne+Wy63z9mvWUVHo2J5FH+uA9HMn3avkirfSMVq2J9biiD8H49njSOUwVuj701KonY6fvFkCeK4awd5szakKghK6BFPLzmOExcr8O76ZJev4a1U8ULlbjqZh1dXJ6JjZBiOv3ubpDJcMZxtJ4ozFOrrkVVSgxt7t5N/MWfluFG9jDEVw5HkCaZVA41wHVcThhBNeKfOqJ0I/4c8QhpAja5GtY+OB7wnNlnAePaLzhESOFa4PJH02SLnux0a5/BbRlYxPo+ZSt+uExcr1LmwHSNn7sY9Px5CfYNJ0vE2HhyZntBtl9bVKa4ycC/oElJPMzNg2Ge78fD8oziaUca5gOMVuLew0IU5PpV1DZKNZUv7ms0MxzLLYGiUVv8WLpTWYE1cNhpNZod9sucIyTvcL/HWAsVKoLQnX/WEQhq4jqevTRBEE2QI+ShyomgCujOVYgh5ohA5Vwvg9irS1yOvsh6J2RU4nSctw5mt4cwfHCdUp66kk3bleeEWU1XfYP0/KbeC/yAeDp4vkRUiFH+hDIM/3okPNp4GAIQEOZfcYvh8svksHvr5CJYcyrLZn1FcjQUHMgQN1DFf7sV//ziFv07lS5aRaEJTc+40gtiAWKARyN8FglATMoQ0gPrrCCmHFkYclY4XF4sGUjt7kUOyBDcTYXjrgymlXLFjaozNSraxUdqNSPEAujoth7c8Nx8Hbpm9O7R2+Vwx/rUsHiYzw/KjFwDwG33ce7F41hYdavI8zdyaYnPsrV/twyebz+K73cKLYQPAsawyp/ulEGjJEpTCdk0tbdWhEolttIhvSk0QhAUyhDSAKyN/Ymeo9U3RXGIBEaTVrZvps92ua/qUWuAq60aeECs5SDLM3CrBNUxmrqeKI4GEl0uOB6uUkwCi6VzJpzol7gJ/sggLfHfhozquV1E6G6WWkBvS66tGkpJ4ZYbQpUIbTWacL6qidiD8EjKEAgAlOy9PLy7aIth5efzrCDnfb49IxJDmEFsrx5dDauSsA3QmT4/lR7JsFHxhzw//HrE5Zny46yG0MYREhLGXW8yYEWp7NRUYk51QSgyWkL7lj8Mjrt+RlutCzXfLNsTde7Xw8qoE3Pb1fqw8dtFrMhCEWpAhpAFc6d/EOkUmYX6EVgniaFI920U47LedBSJWD+J428vluI6Qew3mzfZeG5+DExedewycIcdrced3B/D+xtPYmJhr3WabJASc7UJXUXCQwMlzxC3fzFx/N82MYdPJPBxOL5F5njy5pFJZ14AbP92F139LlH2uM3zZmHcFNe7X12rQ0/L6Wv14A0sdbb2UVGb+/gzRcxpNZnzy1xn8nVIoeuyfJ/MkHUcQakKGkAZQpUMO4NA4uXHyYl4u0QVVpQglA/FYevlle2I08VhmGd5YcxL3zT3sRA7n1zALGDLOOJ2n5z3H1TlL4nPGBLY7eY64ii73HuUaH+nF1XhlVQKm/BIr6zyhcrjb/7kgVnamuA0JuSitMWJ9QrMxylcPchV9TxrzXGl/3HPecwWL4O47q7UwJq443dqG8+wXlldjt2KDp0TTQhVI6a/WxOdgwcFMPL0kzulxhfp6vLwqAU8vidPcs0oEFmQI+Sje6jY8YQeJersY//+u4v7kdzcVFpHfYsdL2e+J5+V8UbXb1zBzpwVJFDq/sq75FJk36kq9uBseyp36JKZY2O/Nq3BtIVupWfOSc/X8BwpeV1oN+kr67C+3p3qpZEfcTZ/tTdWSP2S3me5tnXv6Aft7IUXZG9i3o5RnMq+iTvwgABW1zdkzyQ4ivAktqKoBVFkjQaXrq51BzR41jAJ7XPUAyCnD6fkyG6jpeN2l/92/nlIoUa4r4Vkl1c1JAWxDQsVD0FxKn+3Kgqo2xnvzD/v5NWI8t8z5KKsQUjxCkhHxSCrVRSw+lImyGiPeGN9PmQsGGFpRLhljmL0jFfq6Rptt/oKn1hHSQpVJkUHq68+dm9toZgj1tcm6hN9AhpAmkN/DaaFT9BbuLjhqj6cTQNhjZkBWSY31t6w1onjqwluPhhJZyWzD/pS7E+EkAo7bREMl3ZSFa/y4ZIg4QTB9uMB2NxPzudQP1TeYUGNoRJuIUDDGEBLsGJjAGPDhn2cAAPcO6obLO8pLM+5rqDMYpo2PxOH0Uvy4J130OOdZ4xQUiJCMQ7SChIbIKq2VdG3ugMmGxFw8NLSHDMkIQjkoNE4DqPIRVEmh9LjJIDpfhscQkHvvInOKxBTjitoGrInLFi9HgMq6Btwyey9ni/IPhCcUCde8K7Z1604igaaTmtrwm11p2HQyT8rhshH0eDhLlsD53yTjHis54SPuINQ2js+7+0O+vOmzOf+P+XIPhnyyC9d/tAO3zN7L6xXjyvvFNu2Eq3kSl0LjOA+nVoyHkmqDwzb+8F1lBK5vMOH9DcnYk1rk9DilvFKeq2bvN6gUCeo4Cy47r+PmZ3XHaUqYQHgPMoQ0gEvdm6jXQLqyJeuD4AFLSFQakdAceRcTX5BTSqjPf/84hbTCKsnHO0NWsgTe0DjHbZ6IaFTCI+Suh4ShaZ2bb3adw6+xzalehbOmyS9v7t50fLvL+aKizjCbpb+bb6456XI5NmV6MDROjEJ9k2JcVd+InPI6mzleFrjG0bbTBS6U4vtoxaOjClL6LRc7rSWHs7D86AU8tfi4S+cTzThkNHXze7oxMRd7Lxmo3Eg4pT3jhHMaTGbsPlsIfb0yA22+DhlCGkDteOmUAj2Scyt5953N12PoJ7uw/EiWpGv5QhSvs+o0mRka7eKBuKOo8/aLh3AIUahvmsjOv6Akw4xNpzF3r/tZqcQVJD6vlvpwn+PyGqOAt05Y9opaI974/STnWGBPShH+u+Ykao2NgufZXh8oqXIcgbZvc3eoqG3AnF1pKLdbsFSI11Yn4MF5R6y/uUaZmAIQm1kmS7YGgfsUMgR3n7UfNZf3pPB6UGUqsHzHH80oddiWWlCFl1aeQEZxjcM+LXHofAkeWxiL7DJpIUL+CPep4GtfNbMISp2srxTqriPkpofc3fLtCpVisOgEDJyc8lq8ujoRT14yULnPBRlCnmXOzjQ8szQOjy885m1RNAEZQj6K6Po5nN0vrDiBu74/iBqDozL5v7WnUFpjxPsbT0sq1/PJEpzPgZEzn6aoqh593tmCK97dajPizL2jL7alOtSTEnecVliNJYezFAnzcTdkRi1KOYbBoI93Ytq6JNFzuB/an/amI6PEVsl9aslxrInPwU97mw3UFUcv4P9+OMh/PYE3o8EkbY6Qvr5B8kC0xeiobzAJDjQAwIbEPJzNb87G1shJjcct/mROpXpKlcBlSyUacza4mSxCyuUOpzsaQg/MO4y/TuVr2kPEGMM/F8TiwLkSWWsr8YaKyXwULpTW2BjCWtEt+QeG5F3Dr71jElGyBoyNZqyNz+H1xroLN5ScOwBTZtfX2HqEFBfDI5h9VPC1J3IAAInZFbz78yrqsOlknuxkPr4KGUIaQOxR25dWjLu+P4AznPVSqnmMGjHKax2VnkYBBVEIT9hB8j6S0g2lV1YlWP8v0DenIba/p7P5eoyfsw9r43OwJi5btrLIZ3TUN0hfn8XdrsdboXH2rD4ub95UfqVtamjubXDTRr+3IRmncoQND777bzTze0q4CtZfp/Jw7YwdvEo4/7lNPLX4OO76/iCMjdK8TlyjzN7wEfowSUXouyV9xFXCcSKhmfynuP9BraqX3+d5mr9O5Vv/j7tQ7ta1GKRnFdyWXIAxX+7FJ5vPcs7XrhIjNaTXFXwhakEq3Ex7SjJ/fzreWHMS47/e7/Q4+yaprGvAqmMXnXrDhTxCDsdxWkpLWQSTciptloL482Qe4nne5ZdWnsC4r/fJ+rYrxZk8vTXEUA1u+XIvXlmVgBVHL6hWhpYgQ0ghUgr0KKh0XOPjxMVy1BmdvyhifcATi44hOVePZ5c2uZT/iM/BqM/3CB5vbDTzen/4yjmTL2/dEE9/ZPhkzpUR+sBVBridm2UkZ09qEVILqmzOeWVVAtIKq/HGmpP47x+nZCtgfHUkZ5FXxphTBUhsjtDCg5k4cK7YTiZHqXadKcTU3xMlh52pAddotM+e6srH8UBaCTYk5jps5/MI/RGfg0Pnm42e9zYkyy4PAI7whHFZ4HtWnY3aN3LaPf6CtLC4r3ek4ottKQDcnwv0jcy5T0qEtKitAq0+dhHHs+SFGLrKofMlLp03nyck9511Sbjx012SQjCXHM502KYh3dIBuaJZ7uXvlEL8LpKYRqr3W8PVY+Whn5tDapVoz23JBfjwz9PYndKkRFfJHFA1NJoxbV2S5DT+gssW2H3ftOJ5KK024B8/HMRtX+8D0GRwvLwqAff/5LhI+F+n8pFZUoMD5/jf+TqjCRsTc1HBMwDtDLPZ+fcfAO787gCeXHxckbX7+DBe+kYJ3Zu/QYaQAiw9nIU7vjmA4TN3O+y7b+5hPLbQdhX4+AvluHX2XuvvyroGbE3KhxgVdU0T28QmUF/53lb8+9cTDtvV+DA2mMxIK6xCRa0Rs7enuvxiphZU4Z31SSjU14uubv/WH6es/8tbR8c2o1JKgR5PLT6OFDtDSO7HwaEUnu9wkIxsTntSi3H1+9uwNj4H3+92VEyXiczn2ppcgMcWHsOMTc3hjnwyPbssDutO5GLeXtfnRYmRXVaLjOLmZ8LZrQfZCbmZ806YBDw69ry19hR2nnHMQMQ3R8jdRAT6ugb8y4lC8HdKIW6a9bfDdq4hZP+9s9RBQWU97v/pCMSoM5rw3d/nMXdvOoqq6gUrWKqewf3wuepFdHYe37upVHiJvr4B9849hIUHm42CY5lleHtdks0cLQv1DSa8tyEJ++0GDdzBXgnfxXkW4y+UY8ovR7HkUCYe/vkIzhc19ztcg9zCkYxSlNYYsfp4NrYl5+PERWEPk/27A7iu6CsxOs+9BN/zwL/gqni5Ty+Jw1t/nEJmibrzxObtk94nasXgzCiuFg11e2FFPBYfykLCxQrrtrisMqehvXw483YKeYRsQ+Zsg5idDaiYzQw/7jmPoxmlKKk24Eh6Ke/zU15jxJRfjmLdpbAvV+BGHmSV1Nh8Ixhj1kHtdM43TWj5ow//PG0zJ0oKjDHc+d0BTPhmv6R+UYn3oLzGKNmrlVtRh3fXJ6lmgHkLWkdIAaZvcj6/xr7TeHxhLGrsvET//vUEsmZNcnodKaOvzkY+3c7IxRjK7dL5vvjrCezgfOx/2peO9M/ulH3th34+gsq6BqTk63GC00mLiVzf4FxBFlw/hTGcKxR4mV2spqMZpVh6OIvX+8D9OJgYQ5CIb81oMuMNAUV9zq40PD+mzyVRhYVdcjiruXwnZRXo61HfYELLFsFOZXKF0V84ei5LOel0GWMwmxmCgnQOSuTGxOb01wd5FEU5NJoZGGPYklSAq7pEog/PujTVPJ6/w+kl+OHv8/jkngEO+777+7zNs89l1Od/I6ecXynhPh/vrLedRxVy6at6UcJEe2anTNQZTYJPQ6He0VstRrBOh0aRF7BR4GNtbDTjcDpnNPHSYXyHKzEaXN9gwne7ziHhYgUSLlbgmVG9Adiuz2XPL/szsOLoRcH9rmCvFD27LA6/PjsMN13RwTqqbAm9fH55PHa/cYvoNfemFuHzbU39utA3gi9JhisGzem8Sjy28BjeGH8l/jmsl+jxqQVVmLX1LKbe3g8Du0dLLscS9hccJG2AyH5XabUBvTu0klwe7zWdlDdrawp0gLWflYLZzHA4vRT9u0ahbatQt2Tjw1lfX1ZjxK1fNXkxxPQIex64NEiQOfNOhz64qr4RI3kGeKViYwjZzQni1r+zca4/T+Xhy+1Nc2tbh4Wg2tCI+Y8Nwfj+nW2Om7MrDYfTS3E4vRT3De4ueL2M4mr8ciAT/7mlD3q0i7DZx5Xxru8P2kxBeP23RGxIzMOuqTfbJPURGvixfL8SsyuwNSkff53Kx+cPXIuM4mq0jQh1KBtoCoW0DMwWVxsQE9VS8D4AeSH3Qgz6eCeiw1vg5PTxosc+8NNh5FfWY3NSPhI/ED/eVyCPkBewN4LsWXYkC59tOQvGbDOcSfmunXUS6ua+IeS4zV4RFFJq6owm/OfXePwRzz9aU3nJ28U1gqTw6MJYh9ARJvA/FzNjgh2YlFriC1v6cU86dtll4XplVQI2JuY6jIa5A7eKpV7K2Sj9rrNFuOr9bfg1Vv144PzKegz5ZJf1t76+EZe/swXvbUgCz7qaVsprjfjfH6ewPsG10b7k3Eq8sz4ZL648gXFf7ePNLMWn1E/5JRaH00vxwop4h31/OlmnSMgIApxnsPvnglhkltRI8sZw56MAwMwtKYIjhHd9z59cwhl8XgZ7LAoKl8WHsnDle1ttRkItNcunsHMNdlcZ+skuLOB4goqqxA2/tQIjx/YGRIPJjJdWnuCNl/9xz3n8sj/D+puvzoTmfRXxZDfkQyxzYHmNEcezHEfo7Z9mKSPMb645hbIaI95dLy1M9J8LjmJPajHu++mQ4DF84YIfbEzGoI92WNvpfFEVfjmQwTlCZ5eG2VZ2Z48md19JtUEwk6KYoThza4rT/dbrXKrpNfHZeHRhLAZ9vBMf/nnao/NelPAMCD0eeTwh/0DTnEq+bLNCyRJsy7I160xO6iqrpHlQyGKY/J3iOD/GokOI8dDPR7Dq2EU8udh5xjT7edgbLhk2Cw9m2QwIC60zyB0U+fevJ7A5KR/T1iXh/344xDtAeOliVqQ8Pp9sPiN+EA/215ZSd3VGk3Ueb4VC69tpBTKENMgHG09j/v4MPL7oGK75YLt1u9h70WgyO3157DslofTGvx2/iJM8H29XunXGGL7akYoBM7ZjS1IB3lxzEnVGE2IzSjHx2wOIE4ndz6usw2Vvb8Zlb2/GnJ1pvMcM+ngnVsbyj+xa7vGrHak2C/sxxgQ7MLFEFBdLayWFLQHAppN5eHV1oo2BKDHKSxDupHz7BANC2Cto3NS+lmw+FuXndF4lnlx8DA//fARjZ+/FD3+fw7UztkNNVhy9iAK9sGJoMjP8FpeN139zLZxt5tYUrDrW/IyM5AlZc0aakPfQBS44WXm92tCIfy2L400Bbs/us4U277vS2dScKZt1RhNeWO5oHAqxPiEXjy2MxVXvb3PYJ8UQ4o7A2sMYc3hnb/x0N2qNjU7XycgSaAf7fnLLpdFc+zlkRfp6fLk9FZ9uOYv6BhNqjY1OFTp7gnQ67E0twpRfjko+h48dZ8TbPaukBtd/tEOwD7UgVXlnjCGjuBol1U19h1BWRgBYdcxxTk+h3gB9fSOWHW4yLm/7er/N/LSKWqPDgsi2sjU9nI0mM07lVNhlAW1+cId+sgv/4BkESM6tRG6FfC/pv5bFWdeLs2dLUnM7LD6UhSQn4Wbni6qx43QBFh/KxIiZuyUZMnLsqrisMry2OkHSgIAFoYQyQry0MgHvbzxt9bqazAx7U4tslGqh52ns7L04wklK42xwkM8TJmcwsUhfj9yKOvx7RTzWncixPrPpxTV4YpGtMSRlAEqnsytf4By+QZGDPGG4U39LxIPzDsNkth2cza2otdZfob6eNxzNsibbwXMl+GpHqs17UGNoxCPzj2ARZ4BIPs3XkzM329eg0Dg34SpXSuMwUY01KalC/CYyiZTbKaUWVCGPJ5744PkS/G8tf9rjV1Yn2PwWm8tjud73f9uunVNaY8DD85s+/g/MO4Ie7cIFz/+Ak9b7293ncH2PNrzHvbM+CVOG9QTgmEygwWR2kMHVaJyvdqTyzkERo8rQ/HEwM+b2GiMTvz2A4Ze3k74iN6eDPXiuBI/azVuzUFBZj0nf2SoOs3c4V56UQi9xRM/XsU8Rbs+5omreOX72bEjMwz+Hi4cvuYqQR+hYVhlumb3H+hGWijsTb4W8N7/GXhD0Xiw/ckFwVP/zbcKj/WbGEMx5YWo5Hvyxs/cir6IOse+Ms/Hs1xpNGPzxTqf3YI9OB1nzB4QQaodFBzNxKqcSI/u0t2aT+3b3OUSHt0BMVEtMuraLzfFia21N/PYAVj83HB/9dYa3Pb7Z1dxPSM1Yl1lag5t5Rsfftku9P2PTGRsl8dfYCzA0mDBlQVM/9tzo3nh30jW8ZXDngFbUGrHqWLbT9udSZWdI7zhTiITsChx/9zbrNsv3xv51cTagZpmMb+Httafw2/MjJMkkBUu424bEPMmhchsT8rA3rQhfPnCdrLLKao24DK2w9HAWPvrL1kPB/c5y6ye/st4mLFjIW2kyM8EMgw0mM45nlmFwr7Zo2SKY1x6pM5pw42fNoX1bk20HDfaliScVskdnJ68OTYOTuRV1tuGaInOFLaxLaErusz+tGNdywkvv/+kI/juhH14cewWGXbqH2HfGOYTL/bQ33fo892gXgYeG9gAALDtyAUczynA0owxPXwoVdgdvZJ71FGQIuUFBZb3DWikbeTJWKQUDc1BSuby7Phk3XtZOcP/rvyfir5dHo8FkxoRv+NNmOluscLNdKE6/9xxHd+3Zm+o4AmKfRS+7TPpIw1NLhBWHOqMJW5PzYeB4TD7flsI7kmI/+iIVe4NKKlN+aTY8TIzhf2tPOTlanLP5eqdhkPZwO3hnxjtfwg9P4W7q6ECELwmAM6JahuC350dg4rcHRI+tcxJ/LtcIUgtnIVw/OZnw/pOTBCEmMwN3ulwoJ2bTMnK/NbkAN3D6Wu7SBlJxNUnE3tQi6HQ6jLmyI87m6/G1gJfH4mGxDyOyKKuTrrVVkOMvlDvNuHY2X49xX++1jqgLlQc0PR93fLMfXz3kXKm2/6YIYW94rTuRi3Unmr+1vxzItBpCzvr1Kb/EysqU+uxSx2QoxRJDGuWEUJTWGPHEomMY3LMtXr2tb9PpdhYAY8DMrWdxMrsCy58ZhhbBQWg0maHT6RRRUt+69E26olOkrPP+s+IEVv9rODbyhApL9dzwvQqVdQ0Y9xX/81ZWY8Snm89iyeEsTBrYBT/+czDvdQtkzo18+Gfx/nRbcgGCOHFveRX1eGxhLGIzy7Dg8aG47ZoYmM2MN9uss/frqSXHEW43T/fL7akY0K3ZOPpsy1l8+8ggm2O4Rj13gJU/czDDhdJal6J7pIRK+ypkCLkBX5reV1cn8h67MvYixl3dSXTymzOk9CnHnISaJefqUVFrtJmgak+QwD5XP9oteCZ+3D7H+doFrvLRX6cdwjDs51FYeHtdEp6/+XJV5BBjbXyO5LVqlCI0uKldzS4agIR/EBIchK7Rwh5Yf8LVOPb4C+W46YoO1t8hwY4vTGTLEJswFDkhcRb0LqyLFJtRavUijbuqkzUNshJklNTYDGJzM2NZEDKC+EgpqHI6cKc0s7amYFjvdk7H9OUuFyE2PwvgeITstpsZ8O76JLQIDsKM/+uP03mVOHGhHLf06+RwjfNF1ThfVI19acW4vGMrDOvdDu1bh9mWA4af9zXNo/o7pQi3XR2DMV/uRViLIEleHKlhcmU18gY5CvT1eHrpcbQKdVQnpb4WfAYTN4zNHu5zvzkpHz9CfO61FKRkjLVfV5Dr2fpmdxpuuyYG8QIZHm29YnXoYtcX8w0+ccP3Nibm4aO7HRP3WOBWI3fOa2ZJDc4VVuFweqlgKHJZjRHt7JJ8ZJXWYtbWFDw6vKdgdjx/gAwhN5DzYLyzPgld/26Jw9PGuVyeEmt26KBzGhYmdE98i7FKwZMTRvli0YU4mV3hdFFONfnwT9cmOLpDfmU9PvnrDNbE56BLtOvGOOHbBAfpoKOZoU5ZcfSC1RA6kl5qs0iphZdWJuAuTmiZEn2zFCwhxQAUNYIAOEQ3PMfjDdEy8/alY96+dDzLEwZUVd+AyJYtBM+Vmz6aD/vR/vTiavx6ae7quKs74bGFFoXWeZbZly8t/H32oztstnPnhpZUG/B/Pxy0ztt4fnlzW73+W6LDNdOLqzHuq30O2/lwZeQ/o7gGHSPDHLZzv//Ows5O5+mRV1GHrm2aDQM5Y68JF8sdQtaXHclyaeF5d0jO1aPa0IgGgcW1ud1ERW2DgyEkBSnrigG2S07cO/eQ6MDQ4I93YsHjQzHyivbWbRYDfUtSPqb/gz/01B8gQ0gmjDH8eSofV8a0lhRPykUo+4pUlFhu45bZe7DppVG8+55ewr9AV2xGqaz4/s2n8q3x51pZKI0POWtF+DprONn6pGbXIfyPkCC5vVbgwe2zJjtJZMD1Ntc7GY3+cnuqqqvAu4qYl19sLptW4butgTN24JVbrxA8x5WsihaKqw34YGOyQwgiNynKpzzGtBhXf2Aben7v3OZFPe1DQrmek/UJtpEqf8TnyJqT6mr/wBcyaJPhVCQga+Ssv63zmcyXljuQCrduAOBcYZXN/GJPklNeixCBFKjchE2GRrOkedb2jHcSUcOtY+6UP6ne8WeXxaFfjGNo5MWyWjzjYwMjciBDSCYHzpXglUsjNpe7sI5BvJOFyKb+lqjK+gNcymsbBFM38qWkBGxHIaXw4soT1vhzV0JGCIJQh5Bgx/WaCFt2nClEtaERrcOkfx7Fklvwpbf2NjvPyk/64st85+L8TjEso+b2LDrUnK3LftFuT/LmmpM289zEULJ/uP3rffjrlVHo1b6VjUfLGTWGRoyfs9+tLGVSU9OrwR3fHLCuY+aMe348hIhQ+Wv3GZ0kNSmrMeLK97birQn9BNPGi5EqkBXRn6EgCZkkc7K2uTJiZllUj491Cbk2q6L7Mo0mM55dGqfIGiEEQShDSJDYUr4E0JQm2d95Xkb6c1+Ca4CoxbR17iW78TTOlGd7lBwnqTI04u1LWWjFFj+3sOtsodupmvcoHDYqF6l6XK0C85q4rDqWDWOjGZ9sPovlPOueEfyQR4hQhSve3eptEQiCsCOzpMavs/8oxeH0Ulz29mZvi0FoFDnzUX2NSoUXyzySUYr4C+WSQ7LDQtwfn1/gJwPKhGcgQ0gmFOlFEIQvQ3YQQRBCrEtQfgkQZ5EwXBhj2JDgmIabINSEDCGZHDrv+qKABEEQBEEQhCP3zD2Mk7SeHOFhaI6QTDy9/gtBEISSUGgcQRBahIwgwhuQIUQQhM+z+40x3hbBZyA7iCAIgiCaIEOIIAifp3VYCAZ2i/ZK2YN6tsFRNxZK9jRkBxEEQRBEE2QIEQThE7x759WC+4KDdFjwxFAM6dXWgxI1seTJG9E5uiV2Tb0Z3drIXync01BoHEEQBEE0QYaQDCpqjeIHEQThwMrnhrl03opnhiHjsztxasZ4PD2qN++q1wAQEqRDTFRLvHH7ldZt3z5yPeb+c7BL5Qrx9E2OC+WFBDcZFld0isSIPu0VLU9pPr9/IIXG8TC6bwdvi0AQBEF4ATKEZEArshP2CCnmhC0j+7imaI7q2wFBQTpEtWyB4CAdtr46mve44KCmd3P45e1xS7+OeOSGHrj7+m64c2AXl2Xm44N/XCNYNgCEKrAGhprcP7g79WM8LH/GNUOdIAgi0LiuRxtvi6Ao2v5qa4zo8Ba461plFStCOwS5oB/e0Ns2FEtroVGXd2jlbREUJShIh9uu7uSw3WKMBAXpsOSpGzHr/msFr8ENn9v+2s1uy8S1KyJbqr8iwf7/jnX53JBg/+jyHxveS7FrzXn4OsWupQYzeIxvX+X+wd29LQJBEG7y7cPXe1sERfHYV3Hu3Lno3bs3WrZsiSFDhuDAgQOeKlpR3p0kPE+B8G3eGN9P9jncBXZvuqI9pntQaQmRYLnlVdZ5QBLn3H19VwDAo8N7yjrvlVuv4N0+6/5r8cSIXlj4xFDrNp2MFABzHroeAHDjZe3Qr7OwR++aLlG4d1A30euFcoyLyTcI3+MXDzTJLcY1XaKc7u8UFca7/cT7t2Pvm7dg22uOXrP+XaNwcvp46+/RfTvgKif3bmHcVY5Gp9pc0am16DGv3dbX6f4J/WMctv350ij8dClUsnvbcGx+ZRS+efh63HO9eBt7iqG92iLKzpi+pZ+6beBJr3ZoiPh76m+jzQThb1zmZwOsHllQ9bfffsNrr72GuXPn4qabbsLPP/+MiRMn4syZM+jZU55y5G06R7Xk3R4aHASjyQwA6BrdEv+6+XJ0imqJ//x6wuWyhl/eDkN6tcVrt12JUzkVuP+nI4LH/ndCP+RX1mHq7f3QrlUoDp8vwZtrTmLc1TFYfvSC7LI3vzIKu84UYc6uNOu2a7tHY+Vzw/HAT4eRUlDl0j3xcVXnSKQUVOHBId2xJj7Huv3N8Vdi9o40wfNG9mlvs67TqueGY9bWsziZU2lz3ANDuqNPx9b4fFsKAGBY73aIzSwD0FTHRzPKrHJ0iW6J/Mp6h7J6tY/AgG7R2H22EL/9awTumXsIjAFThvXEr7EXAQBzpwxBVHgItr02Gnd8Y2voXxnTGsVVBpTXNkiqk9uujkGLYB22JhcAAA69fSu+2p6KG3q3w7vrk/DS2CswdXw/MMaQX1mPrpc8UUsPZ+GzLWdhaGx6FqPDW6C+wSCpTCEG9WyDFc8MQ42xEXFZ5ejbqTViM8uw6thFvD3xKnSMDMOba07igcHdMXtHGqoNjdZzVzwzDMMubwcAePfOa9ApsiV6tAvHu+uT8eq4vpi5NcV67MBu0fjkngFoMJlxecfWaNcqlFeeDq3D8OHdA8AYw/hrYhAeGozw0GBB+R8Y0h1/xOdg6dM3on/XKHRoHYYzH01Ay5Cmc7JmTQIAFOnrUWM0oWe7CJwvqsaVMU0K+T+H9URGSQ0mXNMZAPD8zZdjb2oxVv1rOCJCg21CzS7r0Apx792GFsFBmLU1BauONT0bPdtF4KGhPWAa3B092kXghsvaIaOkGn07RWJNXDbuG9wdSw5n4cmRl+G6Hm2w+2whZm1NwdcPXY+yWiPyKuowbV0SXrutL1q2CEbce7chSKfDZ1vOYu2JHJz58A6EhwZb6yw4SAeTuclKn/3gdbjr2i5o2aK5jpY9fSMYA/5OKcKJi+Xo0S4C09YlWffPvG8gJt/Y1Dc/8NNhxF0oF35AADw58jIsOZwFALixdzsczypDWEgQ9v13LGqNJoydvdd6bO8OrZBZUmP9/cyo3ujVPgJ3X9cN0REtAAArYy/infVN8vzy+FA8tywOQFP/+9Hd/dG+dRhi3xmH2dtT0a5VKH7en4E+HVuha5twDL+8PV4cewWScyuRW1GHP0/mob7BjAHdojCwezQyPrsTRpMZLVsEo3/X5kyD3dqEI7dCmYEDbn20iWiBmfcOxL8534JXx/XFiD7t8cj8ozbn/f78CCTlVuLuHw9Zt3VvG46QIB0aL7Vn1qxJKK4y4HB6CSb074z8ynokXCzH1N9PWs/p07EVbuzdDhsS8jCwezR+f34Ebp29Fxmceh/bryP+OawXbrsmBpe9vdnhHi5rH4H6BjOCg3TWerH01RamDOuJ26+OwVNLjkuql76dIpHy8R14d30y1p7Icdj/+/MjcMNlbfHcsjjsOlsk6Zq+yOi+HXDgHC3OTvgef7wwwtsiKI6OMe6YtjoMGzYMgwcPxk8//WTddvXVV+Oee+7BzJkzHY43GAwwGJqVN71ejx49eqCyshJRUc5HSz3BnpQi/LQvHaOu6IC/TuXh7uu74YUxfVBRa0RFXQP6dGwe0awzmhB/oRzZ5bVWRUOnA36cMhij+nZAabURX+9Mw58n89C+VSi+nzwIV3eJQpuIFg6x/O9vSAYAdG0Tjht7t0P/rlE2yo0QjSYz9p8rxvU92qJtRAvUN5gRFATklNeh0cTw/oZkvH77lejWJhxmxpo+vJxR7oultegQGYqIUFu7+Zf9GdidUohfnx2O+gYTXvstEeEtgrHpZB6vHMFBOpz5aALCQoLBGIOh0Qx9XQPatw6zhjY1mMxYE5eDAd2icG33NjCbGXQ64MTFclzRMRJBQUCrS3IEBelQWdeANXHZ+Md1XRFzyUgtqKzH/nPF+L/rutrUz/60YpTWGHDP9d2g0+mQV1GHjpFh2JdajPzKOjw6vBfqGkxIK6zG3ylFeGJEL7RvHQbGGHQ6HRhjqDGa0DosBBW1RpTVGHF5x9YorzEirEWQTf1kFFfj97gcPHJDD2SW1GBU3w6oNZjw56k8zN6RiuG92+PNCVciu6wOn29LwWu39cWGhDyMvKI92kSE4h/XdoFOp0NOeS2Cg3ToEi095I4xhlOXjMGSagPe35CMl8f1Rdc24RjYLRql1Qb0aBeBOqMJoSFBOJZZhpuv7AiTmSEkSAczYyisMuB4ZhnukeAR4VJtaERsRimu6RoFM5MWKlior0ebiBYICxF/ll2BMYbKuga0ieA3rNSksq4Bm07mYcI1MegkMIiiBiYzw/bTBbj1qk6S+4h9acXIKq1FaEiQTehZob4eiw9l4eEbemDs7L1o2SIIDw3tgQn9O+OTzWfRu0ME5jx8PbJKaq0eNsv7zS3bbGaoMjQiMiwECdkViAgNxpUxkTZzrLjyrz2Rg6G92uLyjuIeoiJ9vU0/4gq5FXXYkJCLtSdykFFcgy8euBY7zxSiur4RRzKkL6Q97qpO+OnRIcgoqUZYSDB6Xxo9PXy+BLtTinBdjzaYOKAzWgQHoc7Y1CcUVxnw7zF9EMSR32RmYIwhJDgIx7PK8OC8I5jz8HW4dxB/eFmd0eR0QKCyrgEHz5WgxtiIewd1QwtOH5+UU4nSGoNT71OjyWz9LnD/B5oWwjySUYrnRl9uldlkZnhk/hHc0q8Thl/eDntSivHSrVdYn4llR7LwwcbTeP+ua/Dg0O7IKavDNV2jrPey/1wxRvRpj1/2Z6C4yoD/3XEVosJboLzWiIziGqQWVuGBwd3xR3w23t94GkBTaGpVfSN6d2iF7ycPwpq4bCw9YjsQ2LdTa/z58ijUGU1oe2ngYFtyAV5YEY8Rl7dHQnY56huaBpL+enkUukS3xO6zRRh6WVsczyrD/9Y2fcf/d8dV+NfNl+PQ+RJM/f0k5j06GEMva4fSagM++usM+neNwnOjL4e+rhF3fncA/xnbByFBOvzfdd3QaDbjWGYZhvZqh7AWQdY6MZkZ4i+U40xeJYZd3h7/Wh6HtyZchQ0JuRjQLRrf7j5nvY8Xx/ZB97YRqKpvwLnCarRtFYr5+zOs+8UM+x+nDMbMrWeRUy58TI924WjXKgwnsyvwyA09MKJPe/wRn4Os0hpkl9XhlVuvwHd/n+ctq0PrUJRU2yaYGndVJyx4YihKqo244dNdeG/S1fhk81nB8v87oR9m70jFUyN7Y9GhTMHjLO1u4e7ru2JjYh7aRLRABWfw8csHrsWRjFIM6tkWocE6pBfXoKzGiKr6BuxPK8HVXSJx4mKF9fjbru6Ei2W1eHzEZQgNCcKoKzqgpNqAXWeLsCUpH+eLqhEWEoS37rgKQTrgwz+b2v10nh7/uvlyq3f7rT9O4ZEbeuCWfh0RE9US9849bC3DMmj17KjeWHAw0yrzy7dege//Po/nRvfG0iMXcFOf9tiTWowHhnRHWY0Rg3q0wVc70zCgWxSSc/UOdRLeIhh1DSbr7xsua4vjWeX474R+WHsiBz3bRWBvajEG9WyDhIsVuLxDK2SU1ODjewYgtUCPFUcvWuW769ou2JdWjBsva4f5jw91KMtb6PV6REdHu20bqG4IGY1GREREYM2aNbj33nut21999VUkJiZi3759DufMmDEDH374ocN2rRhCBEEQRGBjGSAhtIvZzHDVB9tgbDTj4P/GonvbCG+LpCol1Qa0bxXq9efSk+9GrbHRYZA20AmUvkkpQ0j1OUIlJSUwmUyIibGN2Y6JiUFBQQHvOdOmTUNlZaX1Lzs7W20xCYIgCEIygaBo+DpBQTpsfPEm/Pav4X5vBAFNYcNaeC49KQMZQY5o4RnwJTz2BNk3jDOLNSwsDGFh/BOCCYIgCIIgpHC1SPIRgiACG9U9Qh06dEBwcLCD96eoqMjBS0QQBEEQBEEQBOEJVDeEQkNDMWTIEOzcudNm+86dOzFy5Ei1iycIgiAIgiAIgnDAI6FxU6dOxWOPPYahQ4dixIgRmD9/Pi5evIgXXnhB0vmWfA56vWNmDIIgCIIgCIIgAgeLTeBuzjePGEIPP/wwSktL8dFHHyE/Px8DBgzAli1b0KuXtNXBq6qa1i3o0aOHmmISBEEQBEEQBOEjVFVVITo6WvxAATyyjpC7mM1m5OXlITIy0uvZMCxrGmVnZ1Mqbz+E2te/ofb1L6g9/RtqX/+G2jcwUKudGWOoqqpC165dERTk+kwfn8g7GBQUhO7d+ReR8xZRUVH04vox1L7+DbWvf0Ht6d9Q+/o31L6BgRrt7I4nyILqyRIIgiAIgiAIgiC0BhlCBEEQBEEQBEEEHGQIySQsLAzTp0+nBV/9FGpf/4ba17+g9vRvqH39G2rfwEDr7ewTyRIIgiAIgiAIgiCUhDxCBEEQBEEQBEEEHGQIEQRBEARBEAQRcJAhRBAEQRAEQRBEwEGGEEEQBEEQBEEQAYdfGEIzZ87EDTfcgMjISHTq1An33HMPUlNTbY5hjGHGjBno2rUrwsPDccstt+D06dPW/WVlZXj55ZfRr18/REREoGfPnnjllVdQWVlpc51PP/0UI0eOREREBNq0aSNZxqSkJIwZMwbh4eHo1q0bPvroI9jnqfjxxx9x9dVXIzw8HP369cOyZcvkV4YfovX2ra+vx5NPPomBAwciJCQE99xzj8Mx+fn5mDJlCvr164egoCC89tprcqvBb/FU+2ZlZeGZZ55B7969ER4ejj59+mD69OkwGo2iMoq9v9S+zfhDex48eBA33XQT2rdvj/DwcFx11VWYM2eOArXj+/hD++7duxc6nc7hLyUlRYEa8m38oX2ffPJJ3vbt37+/AjXkH/hDOwMK6c3MD5gwYQJbvHgxS05OZomJiWzSpEmsZ8+erLq62nrMrFmzWGRkJFu7di1LSkpiDz/8MOvSpQvT6/WMMcaSkpLYfffdxzZt2sTOnz/Pdu/ezfr27cvuv/9+m7I++OAD9vXXX7OpU6ey6OhoSfJVVlaymJgY9sgjj7CkpCS2du1aFhkZyWbPnm09Zu7cuSwyMpKtXr2apaens1WrVrHWrVuzTZs2uV9BPo7W27e6upq98MILbP78+WzChAns7rvvdjgmMzOTvfLKK2zp0qXs+uuvZ6+++qqr1eF3eKp9t27dyp588km2fft2lp6ezjZu3Mg6derE3njjDafySXl/qX2b8Yf2PHHiBFu5ciVLTk5mmZmZbPny5SwiIoL9/PPPCteW7+EP7btnzx4GgKWmprL8/HzrX2Njo8K15Xv4Q/tWVFTYtGt2djZr164dmz59urKV5cP4QzsrpTf7hSFkT1FREQPA9u3bxxhjzGw2s86dO7NZs2ZZj6mvr2fR0dFs3rx5gtf5/fffWWhoKGtoaHDYt3jxYsmK8ty5c1l0dDSrr6+3bps5cybr2rUrM5vNjDHGRowYwd58802b81599VV20003SSojkNBa+3J54okneA0hLmPGjAloRVkMT7SvhS+++IL17t3bqTxS3l8u1L62+Hp7Wrj33nvZo48+6vTagYgvtq/FECovL5dyiwGNL7avPevXr2c6nY5lZWU5vXYg44vtrJTe7BehcfZY3HLt2rUDAGRmZqKgoADjx4+3HhMWFoYxY8bg8OHDTq8TFRWFkJAQt+Q5cuQIxowZY7OY1IQJE5CXl4esrCwAgMFgQMuWLW3OCw8Px7Fjx9DQ0OBW+f6G1tqXUBZPtm9lZaW1HCGkvL+EMP7QngkJCTh8+DDGjBnj9NqBiC+376BBg9ClSxeMGzcOe/bscXrdQMWX29fCwoULcdttt6FXr15Orx3I+GI7K6U3+50hxBjD1KlTMWrUKAwYMAAAUFBQAACIiYmxOTYmJsa6z57S0lJ8/PHHeP75592WqaCggLdsrmwTJkzAggULEB8fD8YY4uLisGjRIjQ0NKCkpMRtGfwFLbYvoRyebN/09HR8//33eOGFF5zKJOX9Jfjx9fbs3r07wsLCMHToULz44ot49tlnnV470PDV9u3SpQvmz5+PtWvXYt26dejXrx/GjRuH/fv3O712oOGr7cslPz8fW7dupXfXCb7azkrpzX5nCL300ks4deoUVq1a5bBPp9PZ/GaMOWwDAL1ej0mTJuGaa67B9OnTZZXfv39/tG7dGq1bt8bEiROdls3d/v7772PixIkYPnw4WrRogbvvvhtPPvkkACA4OFiWDP6MVtuXUAZPtW9eXh7uuOMOPPjggzYfSFffX4IfX2/PAwcOIC4uDvPmzcM333zDex+BjK+2b79+/fDcc89h8ODBGDFiBObOnYtJkyZh9uzZEu88MPDV9uWyZMkStGnThjeJEdGEr7azUnqzX8UEvfzyy9i0aRP279+P7t27W7d37twZQJMV2aVLF+v2oqIiB4uzqqoKd9xxB1q3bo3169ejRYsWsmTYsmWL1SUXHh5uLd/egi4qKgLQbOGGh4dj0aJF+Pnnn1FYWGgdsYqMjESHDh1kyeCvaLV9CWXwVPvm5eVh7NixGDFiBObPn2+zz9X3l3DEH9qzd+/eAICBAweisLAQM2bMwOTJk6VXgh/jD+3LZfjw4VixYoXofQcK/tC+jDEsWrQIjz32GEJDQ2Xdf6Dgy+2smN4sa0aRRjGbzezFF19kXbt2ZWlpabz7O3fuzD7//HPrNoPB4DDpq7Kykg0fPpyNGTOG1dTUOC1TbrKENm3aMIPBYN02a9Ys0cm5N998M5s8ebKkMvwZrbcvF0qWIB9Ptm9OTg7r27cve+SRRyRniJL7/gZ6+/pbe1r46KOPWK9evSSV4c/4a/vef//9bOzYsZLK8Gf8qX0tSTGSkpIkXTuQ8Kd25uKK3uwXhtC///1vFh0dzfbu3WuTMrG2ttZ6zKxZs1h0dDRbt24dS0pKYpMnT7ZJA6jX69mwYcPYwIED2fnz5wVTal64cIElJCSwDz/8kLVu3ZolJCSwhIQEVlVVJShfRUUFi4mJYZMnT2ZJSUls3bp1LCoqyiYNYGpqKlu+fDlLS0tjsbGx7OGHH2bt2rVjmZmZyleYj6H19mWMsdOnT7OEhAT2j3/8g91yyy3W87hYtg0ZMoRNmTKFJSQksNOnTytXUT6Kp9o3NzeXXXHFFezWW29lOTk5Nsc4Q8r7yxi1rwV/aM8ffviBbdq0iaWlpbG0tDS2aNEiFhUVxd59910Vasy38If2nTNnDlu/fj1LS0tjycnJ7O2332YA2Nq1a1WoMd/CH9rXwqOPPsqGDRumYO34D/7QzkrpzX5hCAHg/Vu8eLH1GLPZzKZPn846d+7MwsLC2M0332wzSmAZOeD741bqE088wXvMnj17nMp46tQpNnr0aBYWFsY6d+7MZsyYYWPVnjlzhl1//fUsPDycRUVFsbvvvpulpKQoVUU+jS+0b69evXjPE7sPGmH2XPsuXrxY8BgxxN5fofsIxPb1h/b87rvvWP/+/VlERASLiopigwYNYnPnzmUmk0mxevJV/KF9P//8c9anTx/WsmVL1rZtWzZq1Ci2efNmxerIl/GH9mWsSZEODw9n8+fPV6Re/A1/aGel9GbdpQrRNGazGXl5eYiMjKTJyQRBEARBEAQRwDDGUFVVha5duyIoyI3cb7JNJzv27dvH7rrrLtalSxcGgK1fv95mv8Wi7NKlC2vZsiUbM2YMS05OllVGdna2oEVJf/RHf/RHf/RHf/RHf/RHf4H3l52d7ZYd43bWuJqaGlx33XV46qmncP/99zvs/+KLL/D1119jyZIluPLKK/HJJ5/g9ttvR2pqKiIjIyWVYTkuOzsbUVFR7opMEARBEARBEISPotfr0aNHD8m2hBCKhsbpdDqsX7/emq+dMYauXbvitddew//+9z8ATSvBxsTE4PPPP5e8mKVer0d0dLR1xVqCIAiCIAiCIAITpWwDVRdUzczMREFBAcaPH2/dFhYWhjFjxuDw4cOC5xkMBuj1eps/giAIgiAIgiAIpVDVELIshmS/+FJMTIzDQklcZs6ciejoaOtfjx491BSTIAiCILzKF9tSsOxIlrfFIAiCCChUNYQs2Gd6Y4w5zf42bdo0VFZWWv+ys7PVFpEgCIIgvMKZPD3m7k3HBxtPe1sUgiCIgMLtZAnO6Ny5M4Amz1CXLl2s24uKihy8RFzCwsIQFhampmgEQRAEoQlqjI3eFoEgCCIgUdUj1Lt3b3Tu3Bk7d+60bjMajdi3bx9GjhypZtEEQRAE4RPQ6ngEQRDewW2PUHV1Nc6fP2/9nZmZicTERLRr1w49e/bEa6+9hs8++wx9+/ZF37598dlnnyEiIgJTpkxxt2iCIAiCIAiCIAiXcNsQiouLw9ixY62/p06dCgB44oknsGTJErz11luoq6vDf/7zH5SXl2PYsGHYsWOH23m/CYIgCMIfcDJlliAIglARRdcRUgtaR4ggCILwV+IvlOP+n5qWlMiaNcnL0hAEQWgfn1hHiCAIgiAI55BHiCAIwjuQIUQQBKEQ9Q0mb4tA+CBcO8gHgjQIgiD8BjKECIIgFGD76QJc9f42zN+f7m1RCB+Du64e2UEEQRCegwwhgvAR6htMMJtJS9Iqb/5+EgDw2ZYUL0tC+Bo2HiGvSUEQBBF4kCFEED5AWY0RV72/DZN/OeptUQiCUBjuHCEzuYQIgiA8BhlCBKER0gqrsOBABoyNZod9208XAABiM8s8LZYilFYbsO5EDs2hIQgRyA4iCILwHG6vI0QQvg5jDInZFbgyJhKtwrz3Soyfsx8AYGg048WxV3hNDjV4ZP5RnCuqRmJ2BT66e4C3xVEHyvxFuIiO8/AwCo4jCILwGOQRIgKedSdyce/cw9Z1PLxNYnaFt0VQnHNF1QCALUkFXpZEnAaTGduSC1BWY/S2KESAwA2NI48QQRCE5yBDiAh41iXkAABSCqoUud6xzDLcOnsvDpwrVuR6ADkbPMlPe9Pxwop4zRjGRGBBhhBBEITnIEOICHh0CpsZj8w/goySGjy28Jhi16QFFz3HlqR8AEBmSY2XJSECBRuPkBuhcfUNJjy7NA4rjl5QQCpCCyw4kEEp+QlCRcgQIggRiqrq8cqqBByTmKiAMlz7NsFBrlmdZKsSrsIdjHGn//jteDZ2nS3EexuSJR2flFOJe+cekty3eZrdZwsDekCixtCITzafxWdbUlBRS6G6BKEGZAgRAY+Yt+Xd9cnYdDIPD/18xCPy8IXGKO21IoQJIvcb4UWYG7Fx1YZGWcdPWXAUCRcrPNa3yeFYZhmeWRqHsbP3elsUr9Foan4WDDzZRAmCcB8yhAhChIultbzbz+Tp8clfZ1BZ2+BhibTBwXMluPPbAyjS18s4S/vusiAXPUIE4Sq2oXGeo6penuHkSc7m670tglMOp5fg1dUJ6iZVoa6IIFSH0mcThIvc+d0BAEBpjRFzHr5e3cI0+EF8dGEsAOCFFfFY95+bvCyNcpAdpBwNJjMKKuvRo12Et0XRNDaGEA38AwBattD2OO2UX5r6vyCdTr3+X/vjRgTh82i7pyGcsv10AWZtTYFZhUkpJdUGPPDTYfwRn6P4tbWGzs1QqOTcSoUk8U38Lc10sIvPg7vPkT/yzwWxGP3FHkUzKPojSq0jRI+g58kp548Y8GVO5VTgX8vikFFc7W1RVKeyrgG/H89GZV1gRnYQ5BHyaZ5fHg8AuLZ7NO4c2MVh/84zhQhvEYxRfTvIvvaX21IRd6EccRfK8cCQ7m7LqmXEdAcx5SLQB+06RoZ5WwRFoTlCymGZhL8y9iJG9+3oZWl8A3fSZ/vTXEJfuRdV0517qQr+74dDAIDzxdX4+41bvCOEh3h5VQL2pxVjc1I+lj59o7fFIbwAeYT8AL45GqXVBjy3LA6PLox1yWNUZfDt0RGzmbk16VhWWSqXwxjTdCy/HIXFJ9ZI0Zj+ZTIz1MicBC+VRpMZ3+46h7gsbWYNCxS4tveRjFLEX3CtPciG91+80XcKzY/1J/anNXmr96WR1zpQIUPID+DrHys4bl5f0D2VpNFkxsRvD1jnsIjhtvKgeAXbXvCllQn4+K8zShdCCKA1XfKu7w+i//TtKK02KH7t1cezMWdXGh6Yp72sYUpzNl+PF1eeQLrGw33+8+sJ3P/TETSa5E8W0tqz6xY+cjOB9n0lCH+DDKEAwBXPiNbDEl7/LRFv/H6Sd19KQRVSC6tw6HyppGtpzg6yY/OlBT41i4YflcySGhRUyslq57phrNZovCV71t5U8RHLRpMZa+NzkF0mbSQ3K4DWaHlsYSw2n8rHlF+OelsUSTT64YJkv8ZewHKpi7363+0TBKFByBAKAJT8nlTWNmDp4SyUqDA6LZWSagPWJ+Ri7Ykcj0xwFJsE76kQPEIe5TVGjJ29F8Nn7va2KB5jxdELeGPNSYz+Yo+k411dPFYtNiTk4n9/nHLJGyJGSXVTUo9CvWPfZWg0efU9VqpoLYfGVdU34N31yXh/QzJNTJeIltuTIPwFMoQCAJc+sgId8NTfEzF902k8tfi4WzK5A3dOjhLKi7vZvgLdDNLqt/qCRK+IPWp5Q9MKq7DpZJ6qCvfhdGleUAtiayY1mMw4kl6K+gaTO2JJ7oNe+y0Rv8VlY11CrlvlyaGgsh793tuGl1YmeKxMtdCyJ9/IWRDUKGVxUO3eildwJ5sgQRDCkCHkRWoMjdiTWiTtoyAT7jdEyQ50d0oRACDJR1JGS1E63f3eKp0swZ8dTGrcmtnM8M8FR/H6b4k22zXm7MD4OfvxyqoE7LWblFttaMTxrDJV0uCLIZYq/IttKZj8i2Pdqk25B1Oyrzx2EYB3Q1D5+mhX+gFf8SCQUk8QhFYgQ0gFErMrcP9Ph5Fwsdzpcc8ti8NTi4/jqx2pbpUn9sH0Z8VaDE/cu5brN7OkBquOXVQl1MgdGGOKTf5PKWiaD7bezovATYMtxwujtjJ52m4QYfL8o3hw3hH8ekkh9yRiHqEFBzMBAFuTC9wqR67i68lXykdsB59Hrudd7XZpMJmxISEX+ZV1KpdEEISWIUNIBR6adwTxF8px30+HnR5nCWNZ5QUFSMs0msz417I4fLf7nOixfPqtzSrtEsoT+z6LfZC1bAiNnb0X09YlYdkRiROUVcZikHy+LRVDPtmFNXHZbl9TikdOjrPF5WQJEo+zF9fiXRVbvFiNx0xrXjNvoIV1o/geYfKaqMviQ5l47bdE3Dp7n1vXUTPUVcvfFoLwF8gQUgHjpdF3qZ2Y2n2dr3Wmx7PKseNMIb7emca7X966NZJMIcnXk1KGp/WqjYm5mLHptNPQqjgX1yVRi3n70gEAH/6pXlpwroJrkmEJeWuehTcm64vdq1I1IVZOrbERyV4Kt9WAHcSLa6FxGr0ZeyTcm9r3sj+tBABQd2n+2/miKjy79DiScuQ9h6q+tT727SYIXyTE2wIQ7iPWV7oysujNzymzS4bg7IMofu/qY1+GzkPlWnh1dSKApmx6N/ftiIdu6OHB0uUpkp5U1II4wzxqL3qrBFoUUSmRxPqgu384hHNFzev7eLIutGA68HuE5KOFexFCy7IBwBOLjiO3og67U4qQOXOSt8XRBL5iVxOEO5BHSAOo3ddoUcFyBldZ5hvIl9M5S7l3dzt7+zK8NSr716l8vLX2FPIqtBvzrobXwyYUknP9IJvnSDtzhIQkEZPRn9O0c40gTyM2T8pbuGK8k+IqHfu6yr3Ub2rpNaPwSIJQHzKE/ADROSyuXNOLX1Ru0a4oh9wwHCkfErE7FasKexm9rYt4eo0Od0LJlFbuuZcLsnmOFC2GF3ffGT9cP9OKlpRLLaJY1jgFZFELuXM31cbXnklfk5cgfAUyhDSAu/2baHiYj/Wg3I85nyEk52Pvkaxxdr/d9jC5d7qmP5iqz4ez+eWqR0hddVJIFF97T9XEkyPhmvWi+NkcISmP96HzJdbkQdq9E1vUfG293SV4u3yC8AQ0R8hP4X4QlezLdDr1O0duqIrZA1mfFQ+N8/gsIVs8HU6hJd2ryZhwFMgXPuiiafBVKFNLbecttLAIqVazxhVXGbAhIRf3D+mOdq1C3bqWlLv554JYAMA1XaLcKosgCEIq5BEKAJQMsfCEyhAkEhrHBP7nQ9IcIYWzxkm9XLWhEbvOFMLQaHKrfF9GCVXPNhTS/YK8pRb7QkIHf0SrxqAWFlR9/bdEfLrlLJ5fHuf2teR4PHPKtTvP0ZN4u0fQ6rtBEEpChpAfIPqBUbA39UzohfOQJu4m0TlEHviUuFrCC8vj8eyyOHy6+ayy8nj466mlj6Vg2JmMVlJ7HSEhtGgIeUukQMsax4dLyRIUluHg+aYU08ezyt2+lpy7YWCa6VdSC6rwzJLjgundPfWoaq93IAj/gAwhLeBmDyduBymXfcjjHiGx0DiRrHJKZI1TK1mCRclYfcx2UVFfmytibDRrJlMd91mX+xxYz1NQHjl4Y36Hp0pU8ok2NprxzvokbEsuUOR6WlG47XGpzrx8MztOF2DYZ7sQm1HqsI/7Dop+s5j9b3WzTTrjnwuOYndKkegC6Wrga98CgvBFyBDyAz7dchbvbUgS3O9KXyp0jie+s1yFcHNSvsN+rrLLJ6bc+1U+fbbM8318rO94VjlGzvpb0oKYjgqO++VLMXg8UcNS293X29vb/B6XjZWxF/HCinhFrqfZOUI+mDXuX8vjUag34NGFsU6PE/fka4eSaiOAJgPc02ipHgjCXyFDyE9YcfSi4D5FkyV44FPLLSGtsMrxADkji4pI5BxHj5B31RGllCq5/HkyT/1CXETOyKq3Mm95MsTVgqduVcnnr6zGqNzFNIwvG8wNJr65nc4HsGyOtXtgNO0Y8ZBw5B0iCHUgQ0gLqL2Ao5IL83nEI+R8v22yBJGRRS98POQql942nJRCC3NclPAI0QLH2iY0xP8+W4qtI6TlrsRmAEus39beukPeQE44oRr4y7eJIJzhf18UX0TlDk5Zj5BnEftg8i1CKfcDKtbZi+93/lsu7rYXn1LlCQXJ5PnIEQdsRp1dVCI0rUwSCAlStoG02t6uhcZp9GZgN4Alc14reUO8gy97JQlCKmQI+SncD4ei6bM98J21UWDF9ksYWRRF4QVQvb2ooUcWkeUpRIpHyCHkRYEPrbQ5Qh6ZJeTW2aLhQircg7efVSE8qfhqQcdWah0hjTYnAHn1zJjEtPh+jtDADkEQykGGkJ8iJ3xMy4gpCKIfB6+EE8hDbvuIGn8yy3cFPk+cFrA3swR2iKCuNqnRqiM0hr8pvnL6bTWSqhAEQfBBhpCfwmwtIcXwROiF+ERaOdcSP1jsjmSPsqo+50vd60uTQQNC8CDkCZU1R0jDo+q+izafF0Ab7c3r+XbhOhq4FUHkrP9mv1/Lg3mqSmbTh2m3DgjClyFDSAOo071Jz9DDh1DIjKeVBt5JxNz/xeYISbh5sfAg0Vu2K8PtOUJuJg/zhJGilEdIaVGF7H+N2m22yBwlJ/wbswsvmW3fJ2EQyIP9uZyxOfv9Wn72PSUbJUsgCHVQ3RCaMWMGdDqdzV/nzp3VLjbgUSvbjCe6RTnzm3xpPQql0EJonJZGJ23mEigwR4g+/YSnUWrwgvsuaC181bZf15hwGkWlwA6CIDiEeKKQ/v37Y9euXdbfwcHBnig2oHF3jpBwsgTPhsbxJ0uQ7u2SNCoqSSrpF1C7jlz5ICrueRHxxAmep6wYTgvwdupZX8AX1xEKFFyqMwePkLi32xtNIymHDY1I0HtDEB7AI4ZQSEgIeYGcoEZ/764SqJX+V2xxUCW8I25/cO1D47z8AfdM1jiFrqPMZTjXYwL/S0f19nOx8tRoVq0qWlqVSy345wi5N4CltSp055sUaM8DH+RFIwh18MgcoXPnzqFr167o3bs3HnnkEWRkZDg93mAwQK/X2/z5M2p3b0pe3zOhcc3/94tp7fxYGddyGZmasRb1aKWVe5dD41R+2AVD42RUmqtx8VLrWEgSUnMILq6959zQOO0+UaLvo33WuAB9O1wdzCEIQjqqG0LDhg3DsmXLsH37dvzyyy8oKCjAyJEjUVpaKnjOzJkzER0dbf3r0aOH2mL6HbapSpULjfPMBIpmeS/r0Mr5kWIeIwWyxslF/dA4uVONlYd3IVsvza4RWkCXFlQVJ5DuVcvwPZ/uGjJas4NczeJof26gQnVAEOqguiE0ceJE3H///Rg4cCBuu+02bN68GQCwdOlSwXOmTZuGyspK6192drbaYvodvpwsQQzb++HJKidzhqm7hovy4V0i+zWQXYzPuHapGhWfuyQ90YbSaOHd0CqyFV9VpPAt3JwipDnkLg7K7Ze1/Dyo6a0i44cg1Mcjc4S4tGrVCgMHDsS5c+cEjwkLC0NYWJgHpfI/3O5ABb6oHkmWIDJyyP3w8HkmvB1O4G1lxBP3rLWMVBaEkoRoySOkRPieUlB6XK0gMqAjEe6zK8WjpNPpPKZty/LU2z2WND8G0LY5SBC+i8fXETIYDDh79iy6dOni6aI1ixqdvKtKoBieCKURcfgo7u1SPjRO3vE+OXGYNzTOO0h5HuSlz9amcaBEu5fXGPHfNSdxLLPM/Yv5CZ4Y3HENPwuN4/7v/ehexVCznmXVGUEQLqG6IfTmm29i3759yMzMRGxsLB544AHo9Xo88cQTahcd0MidJ2OPVpRB0QVVRUZSvWMoeXeOkCc+mK7OX7A/S4mwEsFMca4+B9p49FXh820pWBOfg4d+PuLRcuUO9gSa0ic211EqQvPltIA7YatauxdvQHVAEOqgemhcTk4OJk+ejJKSEnTs2BHDhw/H0aNH0atXL7WLJi7ha3OE5MgrnnxISjC69PKk4O4As2hKcNE5QsqE2TgtQ9nLuYWUOWGekNfddlc9yQZjWH3cdr6lZp0hhEvhp7YLqno+UYwzuNLIHUgJNMPYgtdDAql/IAIA1Q2h1atXq10EwYM7GXq8jdjIofh+/v+FcNf7Zf+xUj19tsrXlwKfIqMFpVqobuSlz/YO4mtiudfyd3xzwK3zCfUQmwspFRuPkBY6ChdhYLbvoYbvxVP17JX21HC9E4RSeHyOEOEZ3E6f7cVkCVx4DR3O/3wKudeTJcisIuXX+PFAGQpVrBLXETJ8XR0MEHrGi6sMOHGxXJ5wPHhrTZTUwiqb39lltZod8A3UdWO4uPtuSOn3Pdmdu7WgaoA+D+6GuBMEIQ4ZQgGAr3WfYpFO8kLn3FcGZBs2MtVL+ckSvDNHqLjKgAUHMlBWY+T9KLtiJCshquAcIVezxglsv+HTXbhv7mEcz3Iv0YCaI7tymmD0F3vUE8SH0IIxqNwcIU7Kac11/PzvphBqe7e0mySDH+21J0H4B2QI+SlqrSPkCdydIyQ/NM63EE+4pE6DP7csDp9sPouXVp7grVcp9ahKhkQJHiElhwMOnS/h3S7HAGaMocbQaLtN9BzJl9ccPiy613D3PdZanbvnESIIglAHMoT8FObklzt42mjgV5ydjyx6+6PprXVo1CYxuwIAcDi9VFUZXv8tERPm7HcwFKQgPEdI+jU8MVDce9oW9J++HRnF1ZLPWX7kguxyDI0m6/+twxynhNqOujdVktnMcDyrDA0ms6yyiqsM2J9WDLMCi0z5stHnCkolOOE+uq5mdlQLd5IlBCpeH9D0tVFCgnABMoT8FHdSlXobsTk+Yh8HufeutOKr+rdD3CWkOkopMnwK4PqEXKQWVuGggOdF6vXEQiyF8OS3f2tygdP93PtJLaySZWRkFFej33vbMG3dKQBAWIjz7t5S1Ny95/HgvCOYtzddclkAMOGb/Xh80THedvO1PkgLuFJnNs+8xurcnQQ+6niSFerDFLmKlHI01qAE4SeQIaQB1OjeXFUCLQgpgx4JqxYRWM7IoqRYdBHVV+yW7UtQPQ2yyD0pMCAvQQYeFL7tqnppHiHh0Dhxg5gxhqMZpaioNVq3qd9+zbSNCJV1rhwD9KdLhsyqY00ps018mf44jWa85AGavSMNAPDVzjRZspXVNNXhttPOjTvCEX6/t7uhce73fUoid84eVzYtmwBqprhWa2F0giCaIUPIT/G6S10hROcA8Z0jcr6v48raSUrXg6vhT2Jnca8brGDvJKQU/nkqH4/MP6pIammp9hO3Lcwixpr9NneMXJPIye4mgbAQxFMPcm1Lb722Xl+3hYNLHiENRwLImbNnL7snkiX4WO4EgiAUggwhP8L2I+58Ho0Y3vwoyJnf5IqhZI/ioXEqK32ixoTGFCBn2IvayBE+SGJFCo2aSjGIt18KTSvQ11u3qb8OVLMw4h5NW9wJSRQzXtV8brSmlHPRwto7YssEKHVNB7zUz4uFNAP2IdIafoBUxOvvjbfLJwgPQIaQBlDqWyQte5a7qP/lFJ0DJJ5gm3O+hPAQN2/Jvgj1kyWIGYc8IVAKy/TUkuOOZSjwbJhsPEISDSEb40fAKBKqMr4iVG4/rkEi992UYwg5GlEi1xY5oKLWiE0n81DfYHJ6nFbp26m16DFa0vvcnYcnLTTOc4gNUHEfvyCdDgLjegGL140igvBTyBDSAEr1b0IfTn/rQOUYeUrcutw5I3INAvuj3V5I0b3TJXG+yDHbmUS7xSmN5uZMZdI9Qpz/BcJvhJRCfjtIXfWQq/DZeCN4Qxptt4mFtzmDb46QrVzO9z++6BheWZWAWVtTZJctW2oVOi2phrU3UCOcVcqj4tEFVbmeUB7hbJ4/O7n87BMmGVtb0Au1oN1XhiAUgwwhP0JoAM2l0DiBHtATH045WeP4PvbyvWHuJUtwOF5tj5DYfh+yfO1FNbkQGmdzPYnluFOGu7hjzLgTvsanfMoJCzuVUwkA2HQyz+lxnpyALwcpr4W23h35sgglC9EC4h6h5q06gWPURItPrZbnfBGEv+C4sAThswhNvPa1DlSWx4dv/Q2Z5dmvpeJu1jDV55iI1onKAgggRUkXk63BxFGGJFZkOSfjmxKp011tfqmncQ0huRkJ3VFu+TxCQokbnMGV2WRmeH9jMob0bNu8n+emtKCUS7k/r0kpMqAj/TK+oTiLze2074O1fC8EQfg2ZAj5EVoeDXQV/gVTpXuM5Ko2jHk+UYTcltJC+mw+lHjmTC7Mn3n9t0Tr/9x7Fw6Zc47azS8UoiYla5wcb9If8Tmc6zDe63ONA6mX5r4fW5PzsTL2IlbGXpQslxTUeIQlGUIa6jbdfY/lpqhWG6G5fHz7m36Lh7a6gxpzc5WGpkkRhPpQaJwGUKojlZsxS6uIh37x/9+8TaZXQEbZvPLYneWuR0n0o++Cl8wTSBtxd34Md46Q1NaoqG0QPUeoXD6RxZrP7TlctkPfss51VTkWMqC4W6VPzm+Wudym7u33crZx7rPBZEaNQdoaUWLIedaFjrR9/73z7vAP6LgZGiclWYKX5gjxiWYfGmdzro99w9TAm4ObRfp6zN+fjvIao/jBBOFjkCHkRwinEZbfgWplTQXxZAjOD5By59x75VMG5daF+umXnSM2b0otlPBEcRV2V64nN3OiN1QLOV4d++fb1UxiQl4oW4+QxNA4zgMu9VnnKnFPLT6Oaz/cgeIqg9Nzfj+ejVlbU5xms5NTHdLmCEm/ntq4K4qkZAluliEHWXOEPJAsQUNNLYhYnamNpRmeWHwcn21JwSurE7wgBUGoCxlCPgrf6JBtWJA6seKe+HDahkTw7ef/n+8cufeuSF2pnj5b9Ah1BRDA3XS/gO06Qq7NkeD8L/IcAQKGL3dFexlCSPUEmszixzSXb/vb1To2C5TZMiTY+n+rUGmR0ty7dGXA5OD5EpjMDNuS8wWPYQx4a+0pzNuXjsScCvmF8F5Tu6qvlLBIudfR2v2KeR+FsikCnrkXdz35/oil1s/m6wEAB86VKHr9RpMZxkYZHSJBqAAZQn4EExg+0tbnUBw58oopENLW0uAovgrUlicX5OSDbyTYI9n+3Ky6qvoGTJ5/tPl67mbNstku3SWk9gKbtklN5Bl+rmacE/IIHc0otf7fIlja50CpZ+loZhkOCihW3LZ3FkYnpzZ8aaFhwP2+SGu3K5TMp3lb80b7dYQ0ZtN5EHUGNNVm55lCnM6rdHoMYwy3frUPN362i4whwquQIeSj8E98FjrWl0PjHGUXW3FcrjdMCyvLc3E3K5y37sHdUdsFBzJRxAmXElNc/4jPwaTvDtjKcKnt96QU4UyenrOdH77nRyxU0l3EFEIuSbm2ysSd3x5ASbXzkDI+hAyoHWcKrf83CrmN7LD1mPHsl9h5bD6Vj0cXxqKgst5hn1Svmbw5QuLHchNveBKl1hGSZPxz8KQXRMxDa/+Iqj1fS7FkCR4zOTXwcZLA2Xw9nlsWh0nfHXR6nMnMcLGsFhW1DbhYVuMh6QjCETKEfBTeLlFoNFzBcj3y3ZSh6CthFIhdT/YCqTIryWRmuO3rfZKPF7slb034rmsw4b65h/D5NuEFN521R1W97ci/mCL35pqTOM0xdizXP1dYhaeWHMdHf52xbhdMFiD2/Djf7RK286Ca/6+sc0w88MyS4za/9fWN+HbXOdllSlGKXZkjpET98M0VkiqLLI+QBONqa3KBjCuqi/tJOcSP8docId7QOGXetjqjCZO+O4CZW84qcj1v4otesQultZKOE1vgmSA8BRlCfoSy6wh5zyUk5tERNfJkhsYJlS35HLtTXKm580XVLpwlPles+TiXLi+LrUkFOHGxAj/tTcfTS45jT0qRrPODHOYFyJehorYBO88WOmxvEAi94Fe+nHs8XIHbTkIpq6sNjQ7GUBVPWFhdg0l2+VJC6hpN0m7WpplkVhDf88o3bsCVxdlAhKvtw5VDC3ND+Ps5V/oi514XBzx460Kp7Zv322519Ru2PiEXp/P0+Hl/hiz5vP8UqM/6hBxM+eWo5OxvrtRJeGjzvENnIW8SHdAEoTpkCPkoYguJMpV8QmqtO1FrbMTmU/molplWl0+RtQ0Psd339Y5U/OfXeJtMVLahULKK50VtvUpsXollf1phlcdkAmxH+P5OKcJTdt4MMYLsLCFXFMHJvxzFF9tSHbYbBWKt+EO73JOh+doMx7PKoK+3NW64Sr79e3zGzsPFhytNKWX0VbpHiDunTh6NPC8Y37PpLIOYbfnSJZATkqgF3A2Nk9Ke9l5YNRGaG9e8kfOvG2FyDXKykSiAms+S0l/x1387icPppfh6Z5oCV+MnLKRZrXQ2v8/kY+8j4b+QIaQQfGEtUnFp5I9vm6IeIc/y3oZkvLjyBF5bnSia7EAoKUTzfuFyvvv7PLYkFSA2s4x3vxLZidw1FuWE+/Eagpc2jZ+zv1kmD1hCwfYuHZkE2cl4MrtSsWxRDQLeDuf+IPG2aDSZYTYzLD6UidyKOpt9f57Kx4PzjuDaGTvw/d/nrduFQuMAR68YH/ZNWWc0iY7wcsu8vGMrgWPEy7ZHqH709Q1YcCDD4TipXidn85VcfSaEug2lM5QtP3oBd31/ALVG6UYGn/JuY9SYmWy5tNTv/348G5tPNWcJFJvjysBc/oZpLVueO6j1HS+vVW89IK6c9gMwCw5kYH1C00LPasy/9BQVtUZ8tuUsUgrEB64I7UOGkAIsPpSJ6z7cgeVHL3hVDqHQg+IqA9KLXQu9UpLTeZWIv8BvgKw7kQsA2HW2ULTzF0oTzrdNqK81NPKHF/EeLncdIYHjL5TWILNE+qTQGkMjVh276PQYXoOYLwEAz3H1DSakF1ejsq4BP+9LR35lHc9RTUjJ6mNvyABAVkkNHv75iM02IQXR3ghYcjgLX25P5VVs7I0OMXactp37UVXfgM+2nMWJC+VOz6tvMOF4VpmNIbH2RA5Wxl5ESbUBN3y6C4M+3okP/zzjcO4WjuLHHYFtcLJWEtcrllpQxfv8/h6Xg0cXxFrrZdhnTTJU1BoFn/eGxqYdoSFB2PrqaN5jpCZL4JJsl8zBwrvrk/HJ5uY5GhZFv4GnDL5iVxxtfu6dvX5ydCmp2fpczcwHANlltXh/QzKSc/W498fDks97Zmmcw7atSfl48dcTKK02YOxXe/HCinjxC4kMkniDvIo6vLX2lM33kb9fF24TOXfi6eyAnipOSQNP6qUMjWa8KnPtIG4/wm3T7LJafLL5LF7/7SQYY07XB9M6Mzadxvz9GbjjmwPiB/so/jSgIIa0hSMIp1iUoPc3JOOx4b1kn1/fYMYPf5/DS7f2lXyOs5jyDQm5+HZ384Tqf/96AgAQ+844xES1lC0fF1cdC4wxaxaZ+PduQ/vWYTb7g4N0vAoI/7soFhrGPVL8ZbbxAKgUVdFoMmPMl3sBAGc/usMmjtqeIxml6PfeVnSObsk78VRsEcylhy9gyWFbo5zPSLn/p8M2yQY2JuZhy6ujkV1Wi42JuXh+TB+0CA7C1ztS8R3HoyEEn1f0rbWncIzjfTM0mjFwxg6bY2qNjYgIDeF9tubuTcfcvenW37f064if/jkEN836W1QeLmvic/Dlg9dZf8/amoJfY/mNTK4c13+0EwDw+m1XWrfllNfhnfVJOJldgfJafk8wYwwRYfxt3MgZ/f89Lttmn8UOMpkZJnyzH0IcPF+C1cezMfnGntBfCm86lVMpqPy+vKqpDwjW6XifBQDYfroAl3doLVimBcvpWSU12JNa7LB/yeEsh22xmWVYn5CDIr1jYgSxsL0D50swsHs0WoeFuOXZtA0bEz7OxJjLH8YtSc3GbyonNNUVVh9vejbSi6txobRW8iR0C1rRYyp43hF7b1dQkM5xjhD3fyc3Yzmf7zytcyyzDPP3Z2DG/12D7m0jHParlfhGjpG8MTFP1rW5Xl/uIAd38LGuwWTzvdeLhGjuTytG7w6t0KOdYx15iuIqA9IKqzCyT3ucEhgA8jX+PJmHHWcK8eUD16Jli+bv1c/70rHgYCbWPD8Cl3XgjyDwJ8gQ0gizd6Q5NYQYY6Kx8Za+7TWBFLApBVWSDSElI6mGfLzTZm5Aod7gaAjpdDBduqe0IlsFIqVAj57tIhBxacHHs/nN++WOLIqxN60Id1/fDYwxVNQ2oG2rUNFzDHbeEm47/bQ3HaEhQXhoaHfrtvJaI8JDw0WvKaT42HjEeO4vMbvCYRtfe9pnXDtzadG80V/sAdD0TC5+6gZJRpAQFTwhGPYG79n8Kgzp1VZQQeeyN7UYD9l5mFxByAgSYs4ux5j63+yMGC69p20R3Hc4vXntnoxiWw/hrK0peHN8P3Rt4/z5AIBp65LwyA09bLYJeTNO5jR9uIODhA2hLUkF2JIknjUtp7wOO04X4F/LJXgoOLz+20ne7c8udT6X7Ke96fjpkjF876BuuDIm0rrP8vzXN5hQY2hEdHgL1BhNiA5vcWk/w8KDmbgyJtIukUDTiPSK2AsOBou9h8pkZnhy8TH06dgadUYTsstrsejJGxAWEoQXV55Ay5BgfP3w9U3nqqC3phQ0y2ev9FuoM5rw7oYkVAoY5lIwmZnT8NbyGqOk/tCeIJ64E0tbFFTWY+K3+3H/4O54YuRlnP12x9uda+ljd54pxOu/JWL2g9fhjgGdba4NAJklNYhqGeLwvQEcjQHua+HqvcrF0peV1xqx9t8jnR6r5KOVV1GHSd8dwMu39sWYKzui0WxGZMsWilybG+bJHeQI4TwIxkazzb4vt6dg9b9G8F7vSHopHl90DACQNWuSIjJKpcFkxmMLY3Ft9zZYGXsR1YZGzHt0iKRvldYxmxleXtXk7bu6SyT+c8sV1n0ztzZlfv10y1n88vhQ6/YGkxkmM7MxmvwBMoQUJqdc3qidFPalFeP13xLx+f3X4vZrYgSPE59bosT8F3kczypDqd38Bb7RqKAgAJcGjLiT3c/m6/HGmpPo1T4C+/47FkDTfCIL/KFhjv8XVNYjo6Sa9xgur65OxA2XtcOSw1mYvz8DS5++0WZ/o8kMo8mMV1Yl2mwv1NdbjUxuHVlSSS8+lClQonzqjCb8nVKIQT3aSu6Q7NuN71noxqN8P7VYXsIDe/hGg+1ZeDAD209HSL4X+7V1pDJnZxpeu60vdDodrwfymSXHMeeR61VLCCKF41nleJizqKwY985tDr86lVMBsSk4QTpp85DEkGsEOaOkWvp8hfUJuTa/GRjqG0wY8+UeFOoN6NU+AhdKa3Hwf2PRrU04pm86jWVHmryj7TmK7amcSnyy+SxO8gwa2HuoTlwsx4FzJTjAWfx199kiDOrZxmo4fnzPALQKC3FIBX6htAa92tuOqDLGkFlSg94dWsn2cDWYzQgLcnxPFh3KtIYXW9DXNSAppxIDu0fbbN+TUoS2rUJxfY82Nts3JOTi/iHdIcSgj3di19QxuLxDK15jTAi+98lSwz/vT0d5bQMWHMzEIc4AAQN/H/X88jhcKK3Fny+PQovgIDy3rCmk8IUV8VYlmXva2Nl7AfAr0M4+h4M+3onED25Hmwj1jSGgKeV/SbUBoSFBiOIYJLaJVZQrzzIw8sKKeLQI1qHBxJDy8R2S+uBzhVXoyxmMsIc76CkU/jZy1t/474R+1t/2A0JcTlxsDl8urjIgLqsMt18TgxCJC0Bz4Rr71YZGZDopFwAOnivB0YwyHM1ojmrYfbbQpg9NuFiOQT3bOpxba2zE8iMXML5/Z/TmeFUOnitBaEgQbuzdTrb89gPirhJ/oQxPcr7tfEsYAI4Da2O+2IPy2gYkTr8dYSH+YwzRHCGFGfX5Hpvfjy86ho2JuQJHS+OJRcdQVmO0dvqAuCdEDm+vPYWP/3Kc46AE3NFMC3wTg4MFXu4FB5sMiAultZi+MdkhC42ZMZRWG2zCjfisnBGzdmPKL7HW30X6pkUcj2eVYdUx25H9Qn095l9KvfrRn6dtrnfX9wdxzQfbscsuRfOwz3bjZHYFtiUX8M61yCmXN6fFGXd+dwBPL4nDoI934snFxySdwx3BKqk2ODynQNO8m6kKLyhZJNDBctmSVID5+zME55soxbe7z+GJS51/Zx7P6O6UIvzw93nNLCYsBa73jzFhxcNCcJBOE+mileLV1Ym46v1tKLwUcmfxot7z42EsOpRlNYIA27TjD847wmsEAYDJzpoUCkPmbrYof4vsBjzGfLkXZ/NtPa/f7DqHW7/ah9k7HLMbimEyM3y/+xwemncEdUbTJfkYvtzueK0pC2Lxjx8OYm9qcxr7vIo6PLXkOO758ZDD8c7mCFr4emcqhn66C7MujRgnXCzHP74/iNiMUodjLRlAebMCmhlqDI02dcutJ3sjyPJz++lCpBRUCbYd4DycjDHWXG8iPhaLseAJ9PWNGPrJLlxrFzZ81/fNi5IyMCw/koVp606Jvudmc5OxLWXw05JI5mKZtEHc2+fsx1Ke8Nfm6/HPEeL+X2s02cyr5MsmaYHrpZz03QH8+9cTDu+ZFH7Zn4HrPtyB03lN7fr04uP4xw+Oi75ylzHge0YazczGuOdbEBoAZm9Pw8ytKRj31V7rtsraBjy6MBYP/XzEVmexIzG7AmO+3IPtnLmtL648gf/74RAaTWY0msy80RZSeXd9sk3GSPsBJgs27WdmyKusR12DCVklyg/4exMyhFRmf1oxXl2dqOg10wqreBfgFOvyuJmz9PUNKKisR3ZZLVYfz8bCg5k2HRifqrTsSBbyBF56OczamoKP/zqD/WnF1rhhKSOMS49cwPsbk222ZRTXYMgnu/Dgz0dQ3+D4gWOMwdhodlBm/rc2CecKq/DgvCMOqZW5o1PpxTU4ltU8GsRn2Fm4+8dDeGFFPNIKnSemUFIPFcp+51ho878/70sXTDawTqBD9AR6NzIvSmV/WjEue3uz4P3P359hozz7El/tTMPB8yVOj/GHkA4uO884rhkFNBn79oM7tUZp6y/Ze4RCgvnrjNtlOUvZvON0k4z1DSbsPFNonb/54550wXOEuP3r/fhqZxqOZZVh1bGLeH9DMl4R+b78eTLfqhCXcrxv9XbrUVUZGlFUVY/k3ErBZQy2JBWgrMaIefuaZJ/ySyyScit5vZgDpm/H5lP5vN+S135LRP/p25EtoHwz2Icc27ZJo5kJDpw4sxH+tTweV3+wDRnF1aIeFqG1x5xR32ACY8zJAs7uuXXe33gaq45lY/85x7l5XD7ZfBZjZ+/FvH3S11IyychMOH3TaRzNKMWvsY59JdeLZZshU/h6QkbBucIqVHMUdsvA2q4z8taoA5rCvKoNjdaIEu533SqjmWHA9O247sMdMDSa0ILH62SvLwgtzXAsq2lwgHvfVYbmb5yz9eAsns/nL3neTWaGzafykZRbiYTsCtz302Fc/9FOHEkvxefbUgSNMXvqjCb8eTLPIW1+RW2DtQ24/QJXdq6x6q1F29WCQuNcQCyWWoz6BhPKJCxollFcbRMuYOH13xJ5PQxio0TPLYtD+md3IjhIZx15enxEc3IHi/W/JSkf8TzZtD7YeFpUZnu4awpYiM0sQ2xmGRYezESLYB06RbaUvJ6FffiHxduWcLECV72/DQseH2pjaEz9/aRgprY/T/JPAnVlpFbrcEewhFJJe5s4kQxuhPvICWkKVEx2oT31PAZUbnkd2rdqnnfSNNjC/15ZPMQf/XUGK+3mptUYGmV9S7gG/EcSvfhrT+QgMbscz46+3MbrYq/A/bwvAz9fUpz7dGyF3W/c4vS6721IEl3c9+11p3BLv04O2y1Gzm6hRZeZbQh0dlmdTSKPGZtOCw5KOdPlLYbzrV/tw4jL29vsa+ojm0+etj4Jt9mFohdXGdAx0na+kaXds8tqrfMrO0aGYefrN+Nfy+IRHhqMJU/dgHn7MvDLgQyseWEE+nQUTkoiGP7EuS+x5AIWj8nn21Lw71v6ABBfo+yrHWlIzq3ExpducnqchUcuGb8xkS1t6sk2axw4/ztJeMGzK62wymYJCBt4qif+Qjn09Q0Yy/O8SWUPx3uaVlDNG9Zp7zEWGlxKznWsb65hVd9gRmTLJuNEX99gM3+71mD7Xtl4BllTaC/QtG4e0DR3li8b6MFzJahvMFnbZ8am04JzWy1JYn7c0zwnmNuncfvFvanFuKpzFO91fBHyCMmkvMaIYZ/twlt/nERidoXDqJoU7vr+IEZKyHp161f78P6GZIftzhYdFVu7wt4A445+m81Nxtd/fj3h8JERMrLKa4z4Iz4HhkYTDp0vcXDX9uDJhMOlwcRkp0LmYi/ns8vibEYunKWrDuabxQvhEWBfJjioyQCPv1CumbS6hOcRCkElmmlK4d/0jrz+eyKmLIh1OGbm1hSrEgI0DcTYZ0O0sPlSJjl7IwgA+k/fjsEf71RCbKekF9dg2rokm/7e2ZpO6cU1ootbc1OcC8KEB5ycYW/kPbXkOC5/pzkBiTPPPF//toQnlIqb5fLjv844GIbFVQabtbAA4IZPd+GQndfVUho3XKy4yoDxc/bjWFYZ9qUVo6K2AZ9vS0FZjREztzSllS+tljYvw8JLlya2A/yepb9O5WGDgEc/KacSd37nPNXzrrOFKNDX4/u/zzk9zp5nL4XsF1TW47GFsfiLs2yAWUCRtqfa0Oiw/wjPILAFvrGD+386jKcWHxcN8TQ0mLFJ4JnkprF/dGEsbzmNZrONRyRIp0NZjRF/ncqzRrgIrStps8bSpfsd9fnfGPbZblwsrQVjDAsOZKDK7t1zFjZqv9+C2czw6MJYPLssDiWXnrU18cIJfiyX5UYVNIekFuBoZnN72Gc79XXIIyST1cezUVJtxO9xOfg9LgeRLeVX4fki56FT+ZV1yBMwDo6klwpmE7OMRjnDzITd32bGkFfB72IVmuMy6NJH/M01zRmhXr/tSrx6W1MGPG/kon9e4kRuobWEagzyjVs5mBlwUWYqXHcJbxGM55bF2Uz4JgIPdxe9DQSmrUtCTFQYbr0qRnLq4BdXnhDcl1Fcg8ve3iy4X2rIntKkiaT3fpTHAJSLvUInFTEjTC4z/jyDJ2/qbbPtDEd5XHiQf84Jdy0sC/9cEOuQfMFkZtb5rBa48yNt18/S4cC5YkwXiLAwMQad2XHeF3cQ08wY4rLKkFtRhz4dW8NoMuOllU2G0i39OjpcUyyUzvZeJB9qw8d/nXH4voyfsx9/vTwKA7pFi67PtfhQJp4dfTkYY3h7bRK2nRbOYKmDDo0mMw6cK8HgXm2tWSIB4MttqdYsjhayOAOiZ/L1eIVjVApRWddgTXbE5XhWOVpxlr8I0gFTfjmKlIIq/OeWPhjdt6ODXjhhzn48OLQ7xl7V7K2yeM4syaRu/nIPhvVuJxrubp+llkt5jRHzD2TggSHdUcgJl/sjPgcbE/OcLxtwaSd3sOzg+RIcOFfsoFOF+Nl3hAwhmdgPqEoN6WKMIaOkBpdLyMk+Yqawt4g7CukKJjMTjGk1MSaYUeqR+Y4pi1MFRuXm7EqzGkJaXjONuz4NFylhi+4gdw0cJVjqo/NeCGURcIISdnyxLRUj+3Twthiq8ohIdkK+NPzO+LeUBV9VxBJSJjT4xl3U2F24375GE8Nbf5xyevyvHO/ZrrOFDsl2uIyY+bfoN2jnmULBVPd82+U5guV/tHefLURpDb+H6/nl8Tj09q2i0TN/ncrHs6Mvx/miaqdLEwBN9zNvXzpm70jDgG5R2PTiKOu+dQm5+Prh67Ey9iI6Robh9mtiXPJKAvwJMyrrGmw8PmW1RquH0n7dOwuphVX4ZPNZG8OKzzDkM4LslxcQGkh/fNEx6NCUZfinven49N4B1n2W5CbOKKsxolVYiEP49GMLHRMyeTOzqhqQISQTV5v/g42nsfzoBWu8rreY+jv//CKgyZXKF5usr2/gTZLgbMFH6zUpDIsgNAOFxkkjpaAKV72/zdti+BRbk8XXoFKT3tO24KsHrxOcI/TdbnkhX87gfvsultWKZlz7SWDQjQ8pA3HO1vt6Z32Sze8kmdnvXPlkP7M0TjAddPGlsCyx+WSWrklooJbLuaJqa4KA5Fy9Q4KT80XV1noY3beDqpEQ7653nL4gBHd+rrNMeVx2nZWWGGJ/mq3XT45cQFNEUdMaSeLH+luyBBoflImresTyo00j8s46xBpDo0tzjuRwNKNM0BAS8kTZp/SUSkqBHhnFzsMACUIJru7iPxM31USriTIIQgneWHNSk1EIUpR7tfjHDwfxy37p2eNWH3dt/scxgZAui8pUJxICeu5StlUpmS2LqwzI4IS7XSi1nQtcwpl/pdVw8J/3pYtOk+CjRGBumRK8INGr62/j2+QRkomaLsH+07erdm0p1DWY3A69s7D7bKHNxEOCUJOP7+6PB+Y5hm8StriTmIQgfIFyN9ZX8VfKJSxsrRYWu6ZeJB25ZV6YK3PmbvvaNjpFLOxTC1jmmctFKKRfKbiLxwrhZ3YQeYTkkpBNKX6lQEYQ4Un41nsgCCLwWOJksU/C81g8PHxp6PlQIkkHoS7eSIKlJqQ9yMRZbC5BEOpy36BuvNtDedar8hRLnroBK58b5rXyCYIgtEqt0YTcijrJnjqxuUSE9/EvM4gMIYIgNM4PUwbhwSHd8dM/B+Prh6/HUzdd5nBMaEgQFj91g+eFA9CvcyRG9umAV8f19Ur5hCP2i2WKwU2H620yZ97pbREIQlFumvU3ZkrIXHY6T15iB8JL+JklRIaQDMTy4BOBR6/2tgvGfjd5kJck0TZtIlogMkz+lMStr47GXdd2xZcPXoeJA7sAAD646xoceGss4t67zXpcaHAQxvbrhJPTx1u3HXr7Vvz6rLKemvmPDXHYZgn9eP32KzH19isVLU9pnrZbS0WMp266DFEurJWmNvMedWwHLsueuVHW9U7NmOCOOIrCl7mTIAKBSd8d9LYIhARs18XyfcgQkkGWXWYSgugc1dLm9/9d19VLkvDTvlWot0UAACR+MB59Y1rLPu/KmEiHbTqdDj3a2RqgljlC0eEtkDRjPNI+mYhubcJx0xXO14Lp0DpMljzj+3d22BbGCcsL82KIHgA8NrwXHhraXXD/v26+HACw4cWbJBntwTodWrlgwKpNo8iH2D5N+KSBXZA5807sefMWh2Ov69FGMwvNaskzpQYT+sd4WwSCINykwovJN9SADCEZ9OnY2qmSQQQe7qRtfnxEL9VXaL79GvUVj1s5q2U7wxWHqjMFNTq8BSJbhiA0JAgdI5sNmsiWLQTnDI3u2wGLn2wKoRt/TQyOvTNOvlB2hHrYELJf1d7CO3dehY/vGYAvHriOd/8Hd12DztFNhvv1PdpIMto3J+V7LRFF306tcedAR8MTEPfOcxcFfOXWK/DjPwdDp9OhN2dB6zYRTavRvzfpagWkVYZFTzqGd/53Qj8vSKIOYSHBDl50giB8ixfGeHc9TKXx2Bdu7ty56N27N1q2bIkhQ4bgwIEDnipaUWbed623RXBKkA7Y+frNbl9n/mND8OH/9VdAIt9hvAtGA1exsjD8cv6F5ex5+da+sjqUUVd0sFktWowfpwz2iGLuiqdHCmLPX4vgIBx++1YkfnC7pBH90X07YPkzwzD2qk7ImjUJ8x8f6rCKtoWhvdoiJqrJuBrbr6PNvm8evh7d2oRbf4eFNI/id+Fst+fw27fyzm9yhc/uHYhITsja5R1a4V83O3+Wnh4lLywOAP53x1U25XiKa7tH48d/Dsbcfw5BOx6vZve20pXpTnZeWws/ThmM5A8n4IbLpL2vcnA1JHMYz9ymCf1jMHEAv0EohVahwZj36BA8OfIyl6+hFDodUFnnX6PJBBFo/JsMIfn89ttveO211/Duu+8iISEBo0ePxsSJE3Hx4kVPFK8oQgrXuv+MtPn9rAtKhxCzH7xOdO5Bn46tEKQD3hjfD31jIvHVg7ajwvdczz/6e133aDw+opfNtgeHdMft18Tgbp5zPr7bc8aRsxHr63hGtF8c2/xyzn9sCD69dwBuuKwt1rwwAsMEVr3mhrZ1ayusxAoxum8H/P3GGNx0RXtr5rClT9+IVyRMnA8PDUZeJf+6Lld1jsRzo3tj5XPDcOjtWzH19ivx3eRBmHxDT6tiuuWV0Tj89q346O7+OP3hBOx4/WY8PLQHgKZR5EnXdsHZgirrNUeJhIk5Y9LALtj44k346sHrbDKkfXrvAIzt1+QR4lNYgSZPBABrMoFe7SPQKjSY12BsFRqMthEtcOzdcXhCguIW2bIFIkKlKepC3rsT79+Oz+4diBsvKcRJM8bj9+dHYPMroxH7zjhM/0d/hIUEWRXJewZ1w6G3b8WKZ4Zh19SbbfqE26+OwaPDezokTujWJhxd24Tj5VubtwcH6XD7NTG48bJ2eHR4TwBNc52Ov3sbnh/TFML2/l3X8Hq3pgzriVPTx+PeQd0QGhKElc8Nd3rvf740inf76L5Nz8Qtdsbe7dfEIPGD23HPoG74UsDDJMT9g5u95uEtgnHNpXrv26k12l7ywnDZ+upoXNU5EgufGIqvHrwOCe/fjk0vjbKGRT58Qw/rsb8/PwJfPHAthvRqa33WucREheHN8U195bxHh2DKsJ425wPArPsG4umbemNkn/ZozQn7W8zjjXGVoZe1dbr/ln4dcV2PNoL75z06GECTN+uKTpEIb9FsbH9w1zW4b3BT9sQ5D1+H9++6BhE8IXUdWoch7r3bEP/+7bhjQGf8746rrPuu5AxenHj/dtH7aRHM/9274bK26NlOulEaGhyET+4RHswZf00MTn8ofb4W330T2qZvp9ZeGVwhlOGHKYMEBxB9FR3zQELwYcOGYfDgwfjpp5+s266++mrcc889mDlzpuj5er0e0dHRqKysRFSU91eQv+Ob/UjhKJiLn7wBY6/qhJ1nClGgr8djw5sMC0OjCQsOZGLOzjS0bx2KQn3zisDR4S1w61WdcOJiOS6U1lq3d2sTjhF92uPz+6+F5VnT6XRgjKH3tC0AgGG92yEiNBg92kXg2VGXo1VYMNq1CkWt0eQQz88YQ3GVwToqajYzzNqWgi1J+egYGYavH7re6tVIL65G56iWNte4WFqLp5cexx39O2PsVZ0wuGcb6Osboa9rwOgv9gAA1v9nJKrqG/H4omOidRfeIhj/ndAP7VuHIqWgCo8O74V2EaGoazChdVgIiqrqcSS9FPcN7o7q+kb8eSoPEwd0xg97zmPMlR2x/MgFPDO6N4b3bs/7Mv7w9zlUG0x4e+JVDvt2ny3E9E2n8cvjQ1FcZUCfTq3RsXUY7v/pMJJyK7H23yORnFuJ+fszYDIzTB1/JcCaQmhuuKwdjmeVYfGhLMx+6DqEtwhGWY0BV3RynMNiqfcTF8tx4kIFPt1yFqueG44jGaX4bvc5AED3tuHY/9+x2JCYi6m/n7Sed0f/zrjtmhjc0q+j4PwVxhiMJrONJ8JCg8mM1IIqXNMlCkFBOszdex5fbEtFcJAO6Z/difc3JGP50QtoHRaCakMjbr2qU9PzERmG3SlF1uvcfk0MzhVW4a07rkKb8BYYaWdEFVcZUFFrRN+YSDDGsP9cCQZ0jcLOM4V4e10SBvVsg2u6ROHjuwfYtNP5omp0bxuOlpcUu8TsCtQaGqGvb0BGSQ2ev7kPGGMIUTAcKzm3EttPF+CFMX1cnu9ibDTLTtF9LLMMu88WYlTfDhjaqx3C7ZS2OqMJLVsEWSfHM8Zs/i+uNqBTZEvUX0onO3dvOu4b1A2XcbyQQs/C3ymFeHpJHOY8fB3uHSQczltnNKGizgjGgJGz/rZuj3/vNrS3e/7KaowY/PFO6+92rUJRVmOETgd898ggzNxyFl8/fD2GX94ea+NzcOJiOT66ewCCg3RoMJmtIXbZZbV44/eTCAnW4efHhiCypaNxxKXa0IjD50swpl9Hm/tkjKGsxgijyYxH5h/FY8N74dnRlzu9lhh1RhOu/mAbgCajvMaFxR2Bpsxvoz7fg9yKOpyaMR6RYSF4+OejOJZVhq8evA73DOqGGmMj4rPK8XdKEZYfvYBP7hmAR4f34r3eX6fy8NLKBHRoHYq49xwNl8TsCnz052m8OaEfwlsE49rubWBmzCGsMTajFH+nFuG1cVdiQ2IubuzdDn06tsbyI1k4mVOJyTf2wNLDF/DeXVejU2RL5FbUWb9DH/15Bo+N6IUgnQ6hIUHo3b4V2kS0gE6nwy/7MzB373k8ObI3BvVsg9k7UnH71TH4amcaAGBgt2ikF1fjz5dHoU/H1qiqb8D0jaexLiHXWtcMwJ43b0FMVEskZlfg3yviMXFAFyw6lAkA6BQZhl+fHYbiagOKqwyorGvA4yMuQ1phFcbPsV1QU4zFT92A0Vd0QFV9I6oNjUjMrsDLqxKs+/9+YwzWnsjB2H6dMHdvOvIq6jDj//rj6SXHrQt+zvjHNRh6WTvc9X3zBP/P7h2Ii2W1mLcvHTde1g4rnxuGRxfG2ixS+cKYPrjr2i745UAG/juhH3Q6HYJ0wJq4HBzPKkNueR0K9PWYdufVeH9DMm67uhN2nW3um72NO+8F0DQ/scFkxoM+tgD24J5t0Kdja6yJl7YI6uUdWyGj2L/mlYcE6ZDy8R2Kfp/dQSnbQHVDyGg0IiIiAmvWrMG9995r3f7qq68iMTER+/btczjHYDDAYGg2GvR6PXr06KEZQ6jW2IjTeXoM6tEGeRX16NEu3GmmH0OjCWEhwThfVIWM4hqM6tvBYRQ74WI5QkOC0L9rtNOyucqSt6lvMMFoMiOKo8gYGk1Iya/CkYxS9O3UGseyyhCs0+H/ru8Ksxm4pqv328+e8hojCvT1bs33cYalzRhj0Nc3IrusFjFRLa3zWvanFaNPp9bIr6hD35hIRIc7VwzlUGNoxJakfEwc2MVm9FuO3K6UqcUJ9oQ4+voGmEwMbQW8e8VVBpTXGhEaHITIliEoqjKgbUSode6RP8AYQ0ZJDXq2i0BqQRXOFVVh6eELCA7SwWRmGN23A9Yn5OK+Qd2wN60Yt1zZETHRLZFfUQ9DowlTb++H8NBgmM0MDLZRBFyDkIvZzJyOsprNDPvSitG/a5RgqJ+34esvao2NCG8RDJ1O5zCYUKSvx2/Hs/HwjT3QKbIlTGYmGHEhVj8WKmsbcPB8CcZd3QktWwTD0GhCv/eaDNu7r++K7LJafPHAtbwDWCXVBpukK0IczypDWmEVptzY03q/5TVGRLYMsSqI9vd6sbQWnaLCoK9rQMfIMKf9KmMMZtb03NQYGhERGozfjmejX+dI6HQ6dGgdir9O5WPigM7o3jYCpdUGZJXW4roe0XhvfTJGX9kRr6xKwK6pN+NkdiXe/OMkGGvy6t0/uDuu6hyJ3SlF+H7yIJwrqsb//jiF2Q9dB8YYfjuejeRcPc7k67HyuWHo3zUaLVsEISW/aeC3fetQBOl0YJfuEWhq417tW+Hpxcdxfc82iIlqiTpjIx4bcRlMZobyWiPO5uvRrU3TAJjlO1vfYEJwkA7Hs8rQOiwElXUNyCiuwQ2XtUNdgwmtwoLRLyYSGSU1aBcRipJqA3q2j8CLv57AUzf1RmJ2BW7s3Q6fbTmL52++HEE6HYb0aosvt6fiiZGXITQkCNHhLbAy9iLuHdQNRVUGDOrRBhkl1egS3RT9ERHa9Gw2mpruJUinQ3G1AcZGMxKzK3B9jzZILajCuKs7WdvMbGa4WFaLuAvl6N42HK3DQtCuVSgiW4agorYBYS2CENWyBcJCgmAyMwTpdPjjRA6u6RKFTpFhiApvgT/iczCkV1vEZZVh5BUdkJxbieGXt8cf8Tm46YoOiM0oxb2DuqG8tgHtW4ciKbcSV3Rsjcq6BlzWoRUK9fXo0DoMZ/L0uOGytsgqrUVEaDC+//scXhjTB2fy9GjfOgwf/3UGT910GVoEB+HWqzohtbAKg3q0AWNNz3t6cQ2u79EGF8tq0a1tOExmZn32j2WWQV/XgKjwFuh+aV9wkA5dnYR/exqfMYTy8vLQrVs3HDp0CCNHNoePffbZZ1i6dClSU1MdzpkxYwY+/PBDh+1aMYQIgiAIgtA2jDGM+nwPymuN2PffsTZJVQiC8G2UMoQ85t+yHwFxNto8bdo0VFZWWv+ys7M9ISJBEARBEH6CTqfDny+Pwo7XbyYjiCAIXlSPX+nQoQOCg4NRUFBgs72oqAgxMfxZusLCwhAWRp0WQRAEQRCu065VqGAiF4IgCNU9QqGhoRgyZAh27txps33nzp02oXIEQRAEQRAEQRCewiMzmqdOnYrHHnsMQ4cOxYgRIzB//nxcvHgRL7zwgqTzLdOY9Hq9mmISBEEQBEEQBKFxLDaBu6kOPGIIPfzwwygtLcVHH32E/Px8DBgwAFu2bEGvXvypQu2pqmrKWNKjh+O6EQRBEARBEARBBB5VVVWIjnaecdkZHllHyF3MZjPy8vIQGRnp9dTRllTe2dnZlMHOD6H29W+off0Lak//htrXv6H2DQzUamfGGKqqqtC1a1cEBbk+08cnFvsICgpC9+7CiwJ6g6ioKHpx/RhqX/+G2te/oPb0b6h9/Rtq38BAjXZ2xxNkQRvLwxIEQRAEQRAEQXgQMoQIgiAIgiAIggg4yBCSSVhYGKZPn07rHPkp1L7+DbWvf0Ht6d9Q+/o31L6Bgdbb2SeSJRAEQRAEQRAEQSgJeYQIgiAIgiAIggg4yBAiCIIgCIIgCCLgIEOIIAiCIAiCIIiAgwwhgiAIgiAIgiACDjKECIIgCIIgCIIIOPzCEJo5cyZuuOEGREZGolOnTrjnnnuQmppqcwxjDDNmzEDXrl0RHh6OW265BadPn7buLysrw8svv4x+/fohIiICPXv2xCuvvILKykqb63z66acYOXIkIiIi0KZNG8kyJiUlYcyYMQgPD0e3bt3w0UcfwT5h348//oirr74a4eHh6NevH5YtWya/MvwQrbdvfX09nnzySQwcOBAhISG45557HI7Jz8/HlClT0K9fPwQFBeG1116TWw1+i6faNysrC8888wx69+6N8PBw9OnTB9OnT4fRaBSVUez9pfZtxh/a8+DBg7jpppvQvn17hIeH46qrrsKcOXMUqB3fxx/ad+/evdDpdA5/KSkpCtSQb+MP7fvkk0/ytm///v0VqCH/wB/aGVBIb2Z+wIQJE9jixYtZcnIyS0xMZJMmTWI9e/Zk1dXV1mNmzZrFIiMj2dq1a1lSUhJ7+OGHWZcuXZher2eMMZaUlMTuu+8+tmnTJnb+/Hm2e/du1rdvX3b//ffblPXBBx+wr7/+mk2dOpVFR0dLkq+yspLFxMSwRx55hCUlJbG1a9eyyMhINnv2bOsxc+fOZZGRkWz16tUsPT2drVq1irVu3Zpt2rTJ/QrycbTevtXV1eyFF15g8+fPZxMmTGB33323wzGZmZnslVdeYUuXLmXXX389e/XVV12tDr/DU+27detW9uSTT7Lt27ez9PR0tnHjRtapUyf2xhtvOJVPyvtL7duMP7TniRMn2MqVK1lycjLLzMxky5cvZxEREeznn39WuLZ8D39o3z179jAALDU1leXn51v/GhsbFa4t38Mf2reiosKmXbOzs1m7du3Y9OnTla0sH8Yf2lkpvdkvDCF7ioqKGAC2b98+xhhjZrOZde7cmc2aNct6TH19PYuOjmbz5s0TvM7vv//OQkNDWUNDg8O+xYsXS1aU586dy6Kjo1l9fb1128yZM1nXrl2Z2WxmjDE2YsQI9uabb9qc9+qrr7KbbrpJUhmBhNbal8sTTzzBawhxGTNmTEArymJ4on0tfPHFF6x3795O5ZHy/nKh9rXF19vTwr333sseffRRp9cORHyxfS2GUHl5uZRbDGh8sX3tWb9+PdPpdCwrK8vptQMZX2xnpfRmvwiNs8filmvXrh0AIDMzEwUFBRg/frz1mLCwMIwZMwaHDx92ep2oqCiEhIS4Jc+RI0cwZswYm1V1J0yYgLy8PGRlZQEADAYDWrZsaXNeeHg4jh07hoaGBrfK9ze01r6EsniyfSsrK63lCCHl/SWE8Yf2TEhIwOHDhzFmzBin1w5EfLl9Bw0ahC5dumDcuHHYs2eP0+sGKr7cvhYWLlyI2267Db169XJ67UDGF9tZKb3Z7wwhxhimTp2KUaNGYcCAAQCAgoICAEBMTIzNsTExMdZ99pSWluLjjz/G888/77ZMBQUFvGVzZZswYQIWLFiA+Ph4MMYQFxeHRYsWoaGhASUlJW7L4C9osX0J5fBk+6anp+P777/HCy+84FQmKe8vwY+vt2f37t0RFhaGoUOH4sUXX8Szzz7r9NqBhq+2b5cuXTB//nysXbsW69atQ79+/TBu3Djs37/f6bUDDV9tXy75+fnYunUrvbtO8NV2Vkpv9jtD6KWXXsKpU6ewatUqh306nc7mN2PMYRsA6PV6TJo0Cddccw2mT58uq/z+/fujdevWaN26NSZOnOi0bO72999/HxMnTsTw4cPRokUL3H333XjyyScBAMHBwbJk8Ge02r6EMniqffPy8nDHHXfgwQcftPlAuvr+Evz4enseOHAAcXFxmDdvHr755hve+whkfLV9+/Xrh+eeew6DBw/GiBEjMHfuXEyaNAmzZ8+WeOeBga+2L5clS5agTZs2vEmMiCZ8tZ2V0pv9Kibo5ZdfxqZNm7B//350797dur1z584AmqzILl26WLcXFRU5WJxVVVW444470Lp1a6xfvx4tWrSQJcOWLVusLrnw8HBr+fYWdFFREYBmCzc8PByLFi3Czz//jMLCQuuIVWRkJDp06CBLBn9Fq+1LKIOn2jcvLw9jx47FiBEjMH/+fJt9rr6/hCP+0J69e/cGAAwcOBCFhYWYMWMGJk+eLL0S/Bh/aF8uw4cPx4oVK0TvO1Dwh/ZljGHRokV47LHHEBoaKuv+AwVfbmfF9GZZM4o0itlsZi+++CLr2rUrS0tL493fuXNn9vnnn1u3GQwGh0lflZWVbPjw4WzMmDGspqbGaZlykyW0adOGGQwG67ZZs2aJTs69+eab2eTJkyWV4c9ovX25ULIE+XiyfXNycljfvn3ZI488IjlDlNz3N9Db19/a08JHH33EevXqJakMf8Zf2/f+++9nY8eOlVSGP+NP7WtJipGUlCTp2oGEP7UzF1f0Zr8whP7973+z6OhotnfvXpuUibW1tdZjZs2axaKjo9m6detYUlISmzx5sk0aQL1ez4YNG8YGDhzIzp8/L5hS88KFCywhIYF9+OGHrHXr1iwhIYElJCSwqqoqQfkqKipYTEwMmzx5MktKSmLr1q1jUVFRNmkAU1NT2fLly1laWhqLjY1lDz/8MGvXrh3LzMxUvsJ8DK23L2OMnT59miUkJLB//OMf7JZbbrGex8WybciQIWzKlCksISGBnT59WrmK8lE81b65ubnsiiuuYLfeeivLycmxOcYZUt5fxqh9LfhDe/7www9s06ZNLC0tjaWlpbFFixaxqKgo9u6776pQY76FP7TvnDlz2Pr161laWhpLTk5mb7/9NgPA1q5dq0KN+Rb+0L4WHn30UTZs2DAFa8d/8Id2Vkpv9gtDCADv3+LFi63HmM1mNn36dNa5c2cWFhbGbr75ZptRAsvIAd8ft1KfeOIJ3mP27NnjVMZTp06x0aNHs7CwMNa5c2c2Y8YMG6v2zJkz7Prrr2fh4eEsKiqK3X333SwlJUWpKvJpfKF9e/XqxXue2H3QCLPn2nfx4sWCx4gh9v4K3Ucgtq8/tOd3333H+vfvzyIiIlhUVBQbNGgQmzt3LjOZTIrVk6/iD+37+eefsz59+rCWLVuytm3bslGjRrHNmzcrVke+jD+0L2NNinR4eDibP3++IvXib/hDOyulN+suVYimMZvNyMvLQ2RkJE1OJgiCIAiCIIgAhjGGqqoqdO3aFUFBrud+84lkCXl5eejRo4e3xSAIgiAIgiAIQiNkZ2fbJHqQi08YQpGRkQCabjYqKsrL0hAEQRAEQRAE4S30ej169OhhtRFcxScMIUs4XFRUFBlCBEEQBEEQBEG4PWXG7xZUJQiCIAiCIAiCEIMMIYIgCILwMtM3JuO9DUneFoMgCCKgIEOIIAiCILyIvr4BS49cwIqjF1FabfC2OARBEAEDGUIEQRAE4UXM5uZVLEzaX9GCIAjCbyBDiCAIgiC0AtlBBEEQHoMMIYIgCILwItysR2QHEQRBeA4yhAjCRSpqjVh3Ige1xkZvi+IV6htMOJJeCpOZVDeCcAdu9leKjCMIgvAcZAgRhIs8uzQOU38/iXfXJ3tbFK8w5ss9mPzLUSw+lOltUQjCb2DkEyIIgvAYZAgRAY/JzLDwYCaScytlnRd3oRwAsDExVw2xNE+hvim71aaTeV6WhCD8B/IIEQRBeI4QbwtAEN5mTVw2Pv7rDAAga9Yk2ecHut7SsXWYt0UgCL8h0PsTgiAITyLLI3TZZZdBp9M5/L344ou8x+/du5f3+JSUFEWEJwglOJuv97YIPk1QkE78IB+iSF+PSd8dwK+xF7wtChEgcL1AjFxCBEEQHkOWR+j48eMwmUzW38nJybj99tvx4IMPOj0vNTUVUVFR1t8dO3aUKSZBEFrF3/S22TtScTpPj3fXJ+Ofw3p5WxwiEPCzd4ggCMJXkGUI2Rsws2bNQp8+fTBmzBin53Xq1Alt2rSRLRxBeAJu6lqCqDWaxA8iCAXhJkjwt4EFgiAILeNysgSj0YgVK1bg6aefFlUkBw0ahC5dumDcuHHYs2eP6LUNBgP0er3NHyEPxhim/paI6RsDM6MZ4Un8S3Mjw5gIFJJzKzH8s91YG5/jbVF42ZiYi9iMUm+LQRCEH+OyIbRhwwZUVFTgySefFDymS5cumD9/PtauXYt169ahX79+GDduHPbv3+/02jNnzkR0dLT1r0ePHq6KGbBkl9VhXUIulh65gEaT2dvi+DQVtUY8s+Q4tiTl8+63H8E1NJqw+VQ+ymuMHpCOUBo/m/JE+AC2c4Q8V+6rqxNQoK/HG2tOeq5QiaQU6PHq6kQ8PP+ot0UhCMKPcTlr3MKFCzFx4kR07dpV8Jh+/fqhX79+1t8jRoxAdnY2Zs+ejZtvvlnwvGnTpmHq1KnW33q9nowhmTSa/dv4OV9UBX19Iwb3bKt6WV/vTMPulCLsTimSlFXu651p+HlfBvrFRGL768LPOaFNyA4iPA2z+d9zllCDSbve3JyyOm+LQBBEAOCSIXThwgXs2rUL69atk33u8OHDsWLFCqfHhIWFISyMUvIqhXY/dcLEXyhHWEgQBnSL5t1/29dNXsXYd8YhJqqlqrKUVsvz7Px1sslzlFpYpYY4hMpQaBzhaShTnCP0GhIE4QlcCo1bvHgxOnXqhEmT5K+5kpCQgC5durhSLCEDrjLna9/Yshoj7v/pMO76/qDosdlltW6Xp5UP7tr4HIycuRtn8nxrTpyvPV9iaORxIAIUd94nxhjeXZ+EpYezFJPHW2ilXyYIwr+RbQiZzWYsXrwYTzzxBEJCbB1K06ZNw+OPP279/c0332DDhg04d+4cTp8+jWnTpmHt2rV46aWX3JeccAr3G2JWUFOtbzDh9+PZKNTXK3ZNe4qqmq8dSCOlb6w5ibzKerz2W4K3RQlsSAEjPAy3l3Onvz6SXopfYy9i+qbTEsvVbv+q0/iLuDe1CM8ti0NxlcHbohAE4QayQ+N27dqFixcv4umnn3bYl5+fj4sXL1p/G41GvPnmm8jNzUV4eDj69++PzZs3484773RPasJrfLk9FQsPZqJLdEscmTZO9fIYcz4y6I+jho0ajtsPBNxRwDJLatCjbThCgl3OQ0MEIFzbx+zG66+vb3RfGEISTy4+DgAIDQ7Cj/8c7GVpCIJwFdmG0Pjx4wVH6ZcsWWLz+6233sJbb73lkmCEe3ANBCU9QrvPFgIA8ivV8whxFVExyZW4NaVHHv3ROAskXG2/P0/m4eVVCRhzZUcsffpGZYUi/Bpmly6BgM94ZnMr/C+pA2P/3955h0dRrX/8u+kJKdQ06b1K700EQUFF5XoVrwp2rljRn8K1UJQLlquIgog0AcUGCkovodeEACkkgZCQnpBeN9lyfn8su5nZndmZ2Z3dnd2cz/Psk+zsmTNn5syced/zlkOQX6FGdPNAVzeFQnE4dNqyCeBu3mWOUuLEHE/JxN0sQ5oCEzC42e0liK3pszecygAAHEu7JWNrpEEIwcHkQmSV2B8752kUVqqxaFcSrhdVu7opVrHHIuQuY5kY3OVUdPZ0mEJZtjcFo5YfwabbY5qnU6+li2g3Zagi5KFIsapIqtfJb1olKHFS/egdcYkKK9WY8e1pTP7S+hpcFPtRemyCNY6m3sILm2Mx7jPhhauN7LiYg8vZ5Y5rlEJ4dVs8Np3OxPRvhJOwOB2Wa5zzBj0ljK98uEv2Rq0HKkJrj98AAHy0+6qLW+J4jqQUosf7+7DuxA1XN4XiIqgi5KE426riKJQczOtM5MiO5yiaUkILpRN7s1RS+XM3SjDv18uYvuqUg1qkHK7klAMAahqUN/vLcoyzxyJkd0scS1lNg+iFppV+LkZ0HrxmX1MY2+f9alhM+OMmoPRRuKGKUBPAnccyd267J2PLC9KRL9WjqUU4kFTgsPpdjd5Bs87XbynbTUxOlGzpIy6yCDkTjU6PgR8dxMCPDqJB6znKg4d2V5NBuaMCxVlQRagJ4G6zOsyBSajpfN4TRVVqVNRqZGmPEi6f0rxEpF6TQ8mFGLDkIGJSimRvi1anx+yNF/DiljjRs83uxKJdSRiy9BArrTwfUvvF0crB5exyLNiRgJJq16cYVtozxIcSxhtHUMXIaFdeK/ycMvvL3d5hFArFfaCKUBPA3d4hrBcgh2uc0EuxSq3BsKWH0X/JAXHHk9Q6MfW5icRlB1Jvqec3x6KiToNnNl2QvS06xv1QqWYrvzGpRTiUXCipPqUJzJtOZ6K0pgEbTma6uimSmb7qFLadz8L7fya6uimKfiqZ45y7jddi8WY8WGIMnKw4Vw+9JkqHXnZKU4AqQh4KW5mQsV4Z6xID1wtQ6KV408UZs+wWpJUssd3GHWZo67U6PLPxgkkJo1jiLKVPCZnalBx8L5drnJLPUcWQNnQiztHZca6ZxTV4dM1ph1itKRSKcqGKkBtzMLkQn+1PEYwfcGefc66WC52NVJcKxckObtBdbtBEaBgL01bX04UmuVDare9IlHyuzOfpfEYpjqRIs2IaUfI5MpEa8+aM8Wber5dwIbPMIVZrd8WNRQcKRTSSF1SlKIcXNscCAPpEh2FqvyjWb8wBzP0GM6YLhXTXOC8zFwxvO6UD97t+jod1f7muGYbj8zSAHWvm6lY6B6lnqbhJAEfiJue6dI8he9W5/0xERGiAi1sjH/a8k5zx+JYoNL4wt7wOVWoNekaGuropFIpHQi1CHkBOmXVXMHcWAjld41jfLKUbqS4VcruTuIm8JRtKvb3YlkHb9pOTkup6nL5ebPE81tRrcTS1yKMyaSkRJT+XXGP0rSrHJ5hw6rPLOJYY1zj2rk5YWNvhR7CN0cuP4N4VJ1BQIZwshUKhSIcqQh4A0wWIC4XKqeKwIUbI1rLughIUWyWt76SAyyGKiV8cwxPrzuHvK/ms7XO2xmH2xgv4/ECq09sklNgju7QWb/92GSkFlU5qkeNQcvyMu9zD9sBOCNEETlhm0ptQqnsKxZlQRcgD4JpJdphrnBNkCWkWHcvfVQKudVJxutDPeY2VlUGJ2Ybk/Er8eO6my4Qbvv5Rmnto+e107oeusuM/TlwrBgD8dC7L6v6ukOPnbI3D73E5ePAb919w1dr1+/VCNs6klzivMU0Qe55BJTy/FArFM6GKkIfCFA7dLVkCU145n1lq8buQYmKrS5RiEGiz0k7pVlU93vsjETsv5bnk+My4a1bKXZtrdI3lQA5FUnIVAqeaUlAFgHuyxd3gO9VL2eV4Z/sVzPz+rFPb09QgPP8rBVvbpMRzoYhHyZZiinOgilATwJaBuqJWg/GfxeCTfSmyt0cKAb7eFtsEF1ll/C8qRkhimwTrc/DAai4w/xmfi5e2xKK2wTIzml5PUOyAxSy5LusFDqXVGfApEM6eALC315UoUDlbRFBrdMgornFI3XzPZXapa9PtA/JN2ChZphOj6B9IKsCiXUnQ6tiKt1tOaFEoFLeAKkIOIDmvEpO/PIYDSQVOOR5nimmWW5D0t8gPZzJxs6QW3x5Nt71hMhAe4i9Qwv5kCULSntTLJ7csciSlEG/+csn03bw5b/xyCfuTCvH98QyLfedsjcOQjw/hfIa8SgqXVc5Vlke+o7Iz2ylHkuK7P+S4fFIFYaHizhKs/76Sh6fWn8O4T2Mw4fOjOH29WPZjKFhHcMj9qbQ4HDGteXFLHDadzsT2izlU+aFQKE6BKkIO4OUf45BWWI0Xt8RZLXe9qBof/Z3skOxALDcEG14oWonrPMgJ88jc6bOFapC2grniMJPYnt0UiyzGrDXf+ZfWWN5HB5IN8SjrTtyQrXnW2uAKeNuisBghI/zNdX4jhayXznIbeeWneJy4Voyi22Phjvhc2Y+hbGuJPPUo2S1YSsxeYSV7LGvKWeOMKL19FIq7QhUhByB28cbp35zE+pMZrNl+uWDOBirthSgF7vTZ4mOEpC7c5w7wJgewso9GJ2+Mh6KuKk9jWFmqXNcMRSMkXHl5lPTFfTJKVpDsQWn3I3vcErOsAaO0A06mok6DBTsScO5GicgWUSgUT4QqQi6kpkEHALiYVebQ48jpsuQMmYHZXFssQqyFNOVpknMRSpbAJ/hb2a9BbkXIJktdI4WVavzvQCryK+rsb4uorHHKvxPkaKLcpymUXttVlFTXQ63RSdpHyUqdI+5OeZJvEGQU18gzocQa123eVTZWHr6Gbeez8NhamiSDQmnKUEXIAUh9/2gF1gGy5YBKz9BjHWbGO2u/iqhJVLIEmSUkFwlcW87exNwfL3Kes15hSb+e/yEWXx+5jqfXn7e7Ll7F0O6aHQNvjJBTW2FAyBqiRGtJUaUagz8+hDGfxEjaT4nnYsQRirocNW44lYkJnx/Fe38myNoeMafr6LkLOSZhKBSK+0MVIQWgcYCU6qjZcGfEDDCbq+PQhKScj1KFYXuwdvq7E/KRWljl+DZwbJNyayTkVgAArhXZv0ggf7IE57rG2Y0CG+kq3cHaPX769no/UrMhKtW6Bdj/PJn2kXm9sS9uL/K77Xy23XXZk7zEEYqit5c48Uer0+PktWLRLu8U90K5owLFWUhShBYtWgSVSsX6REZGWt3n2LFjGDx4MAICAtC5c2esWbPGrgZ7Io6Z+bJuVVEy7JlD69YuLmHB3kQR9mL3wCqUxU5AiKhxwgtbSZ5mfEKSrfeBki0HciN0rl4edDE86FREoaRMiQC7PVLn/lx5Jt/EXMeT68/h2Y0XXNgKCoXiKHyk7tCnTx8cOnTI9N3b23KdFyMZGRmYOnUqXnjhBWzduhWnTp3Cyy+/jDZt2mDGjBm2tdgNcPYLV/gloawXohDsGCHrvwvtL0YYEOovpV09ofNv0DqhxQq6KGw3UO4kIW4RIyRwUUW5E8ndMS5SHqw9k7aOr0rWgxxxe8qTjl2+q2bPuOzKx/eXCwZrGNfi3s5E+SOYe9LUJkgolkhWhHx8fAStQEbWrFmD9u3bY8WKFQCAXr16ITY2Fp9//rlHK0JKQEiZsBVnjxmciR4kBQnJ1hTR2C08CCl6grs7/qS5juEqYYUvLS9rBtoNpAjXWC+bjhSg7BXkXXODCk0QyHnFlBYjJBbhiTeFNJRCodiE5Biha9euITo6Gp06dcLjjz+OGzf41yc5c+YMJk+ezNo2ZcoUxMbGQqPR8O5XX1+PyspK1ociDVe7h9kDW4C1ZZZcmgDsKvHofEYpHl97BmkSY3oEX7zOMAgp6J7iTZPNmgxwfIPtvY9cnSyBCnSuQ7ZLL3UxaSdi15IODjgVJavFRugzSaE4HkmK0PDhw7F582bs378f33//PQoKCjBq1CiUlJRwli8oKEBERARrW0REBLRaLYqL+VcOX7ZsGcLCwkyfdu3aSWkmBfYFproaoYX3pLgQyXHujnoXPbnuHM7eKMUT35ulb7XTVc8Z1g9F3VE8LnB6GxUhdxCQHIE7WM3sQdEGIZlgLR0gR3/KeM3sSpbgDCu3Au9/JbaJQvE0JClC9913H2bMmIF+/fph0qRJ2L17NwDghx9+4N3H3B3BKKhYc1NYsGABKioqTJ/sbPsz1ngyQsqC0lInCyFlHSElJksQi3Ftn+LqBkn7KeGclDRTqWcJWMz/7ZiBdgGyXFM7qnBkn0o2ADjCAqBgRcgRV15hehALMf2rlBghV9JET5tCcSqSY4SYNGvWDP369cO1a9c4f4+MjERBQQFrW1FREXx8fNCqVSveev39/eHv729P0ygM3M4iZOc6QoRHMHYWDpe3BD3jnBEjpBzELajqpMaIgG8SyDWucY1t4XrWFKw7SEbJ8VCOSZYgXKlQCUclSxBjoVXKM+vK96eSJpwoFE/FrnWE6uvrcfXqVURFRXH+PnLkSBw8eJC17cCBAxgyZAh8fX3tOTRFAHuFQL7XnzNmVYUtQgKucSxLgLTsRHJgd312KjpN7d3JnyyhEUmucS6Sl13db5wJMGSqW+oltZ41zrYO4ttNCQqSI4RtWSxCcrrG8cXyidrX8dh6ro5sWxMbyikUlyBJEXr77bdx7NgxZGRk4Ny5c/jHP/6ByspKzJo1C4DBpe3pp582lZ8zZw5u3ryJefPm4erVq9iwYQPWr1+Pt99+W96zaOJ8E3MdH/+dzNqm1NlwqQitI8TpFijx3JUgCElBAbkSFHVPEZ5v7Hgh5TRYSbO8sseUKBglP+WuSp8tdE2kXrP0W9V4/88E5JXXWW2P1PNV0jPjTJroaVMoTkWSIpSTk4OZM2eiR48eeOSRR+Dn54ezZ8+iQ4cOAID8/HxkZWWZynfq1Al79uzB0aNHMWDAAHz00UdYuXIlTZ3tANadzGB9Z8++ue9oyhXfxH45WI8hckvsTJbgDKFBSfcUXzYqR6WQd2eKqtR4aUssjqYWAWhacRjKTp8tD6xzdEF/PvbdGWw9m4UXt8Ra/MaewHL9zWZ+O/A1SaipjryrlDTOUiieiqQYoZ9//tnq75s2bbLYNn78eFy8eFFSoyjy4m5CIPPFo+O0CIl3jZPDEiBVfnK0hUnYNdAJKOieEhMT5hzlUByOFMiF2vDatnicvVGK/UmFyFw+jXWvKslq5giUrAbJdemZ5yjP2CftqhkTvyTmWi55wZqwkNgOz74z+fHwR1IhKHlkoDgDu2KEKMqFPTPuXqOpYIyPgIuFp7tgCLaQK+hd5rFeqVdJTLyQEO7mKimWfYkFOHujlPd3pfapOTb3jogdXfX8uypGSDBZghwN4TiWqAVVHZz10d2ecndrL4XiLlBFqAkg5zvEGUKikEsT4fmfsy4Rx2MqCe5gPROOEXKC9UNAAXUmfOuTsFPIu0HHOpg5W+MstrHvfctr5Crhy5mCrxKef0c8T0qb1JE6Oecua+E5NFmCck+bQvEYqCLkobi3RYjxv50WH6nn7g7XypascXKflpIEE76ZY2fHCLnbjG3H+btZ393g1rcLMW5e7vD8W0PKJJHzYbosSynddFHSOEuheCpUEWoCuPO7nTulr3XXOXvWEeIqL2fqX1kQVP4cc9hbVfVYd+IGSmsaFHVPiXGHc3cB11Gws8Yp5xo54hkSU6VyroAcFiF52iEXUiw8Fn2lsHNxFkrrQwrFE6GKkIcitCipkuHLAiZ6f4m+5e6WQlgOd0BbeGFzLD7efRWv/OTY5Cd/X8nDxlMZosuzZ8G5lWSlPQO/XMjCS1tiodbonHbM7hHBVn8XCMdze8QoV0pw7zRts+HqsxMSSHM/czSE9wt3WXuSK8hJUz02hdJUoIqQh+IxrnFcvwtYfITSa1ugcq/MWXK7A4rlUnY5AOB0eolDX9Cv/BSPxX8l4+yNElHl+RRnPet/ZfXru9sTsD+pEAeSC0Xvs+ZYuuzt4IoRupxdjge+PolzIq+/kc1nMjH5y2MoqFBb/Oaoqy/lXhcT3+gqVyROy7dNk0B8X2xDzqsh1VVVyEW6KeBu7253pAlk1acIQBUhD8VRvuLOGDSElDjBGCLG/3JYApQ2UAqnD3dCG5zwgs4qrRVVjr8lyphRtsbN4hpJ5aUmfSiqUkOrMyzGpRPY1/jr4r+SkJBbgcfWnpXkFvrhziSkFVbjk30pktpoD1JuQyVbhLiwqSl2uAU7Gslr2zn5BGxWgh3YTqX1IYXiiVBFqAmgtNlwYYSSAVh/odrjWucOl0oJbXRUG5h95yVSA+WNEbLRIuRMxZdrnSxrCJ0H8/rFZ5Vh2NLD+Ne6c7f35SpvWfflnApJbTKnrsF+dz9rpyn3JI8SXGO5E5zY4Bon8xpqciJ1WQOlLAruysuosC6kUDwSqgh5KKyXqA2DqSutIEIvTCIgCfHFjIg6Nkd5qSnDHb2CvWCMkBu/PJlWC7FXkU8xJjxlXA1r0UuJFh4pxbedzwIAnMswrB3EZRGS25XKUI3zrrUk1zie59Ke8UIuOF18balHorLhTGydmDDfVy4cPU7LgsL60BNxg7uA4mCoIuSh2OseppSXqL2CiWdahMRbBRzXBsfUy7SQeIkcnfhcJVmCl96+dikFe2b5ORUhjrgNe+8frvHGUcKGlJaKyhrnMouQPDFCrP3t2112+CYpeMs72c3P5jXyHChJ0/TZFIrjoYqQB8HnEuZug6lgDJDAC1LqrChrhl4GScjh2bMFmuiUGCEHHYVtEbLBNY65XcFuQiYkzkrbcxpc14DL/cjeKyXHpRZ7WWSPERJfnROwwTWONfYp7Gwk+jQ6u/lKfE8qrAspFI+EKkIeBN+g6W6DqaBrnGB6bGm+5UwBSQmXyh1M9XKl+zWHpQiJFYZ5UmZLzVJlOq74ok5HSkyRuSLJqQhxWoRsahqzVnsrkHAk8ccSE3PmKuVBaELHlnqUPO6LmZhQjmur62aeFNyFFIrHQBUhD4LwfFPsbDgPQutHCFm77PGTd+ZsNm8bhH4Xei87obsr1RqH1CuU2YwLXouQyACYBq1elKBVUl1vSiHOhS0xB1L3EE6WwP+bjsM9UIqwWa/V4WJWmXD2OQfff7YmQ+HrHrZF2LY22YtQ9kvx9Sh3rJcyLqtgez+LRckTHkaU3J8UiqdAFSEPQs/z4nC3oVRIOBN0nRP4XfTBb6O0mFrh9NmO7/EHvzllsc1mH3sGbIuQnVnjWJMB3PsWVqrRd+F+vPXrZcHjjPkkBg+tOoXYzFKedki/7lLvLWJHrBO3RahxW2FlvdX93/zlEh5ZfVpwPSM5Jl5cJvi6bLC0PqFjSy1SM7M5GqkxQhRzV18KheIIqCLkQfDPjLvXECocA2Sfxcga7mA9U4JFyFEwFSGx9y3hERfEZKnaevYmGnR67IjPFTxOncaQFvpwSpGodolBqvIo5f40v/dtsbYx2ZNQAEB4YVdn3n6S7nUxrnEKEjftT1qhnHMBxD2PprJQjuCvlPTZCutOCsVjoIqQB8EXO+PWA6gNLiNSXSqYwqgSLpWQuCbURqUJQFLQSew783J8+/DpAFzKAdMSxSWM2rtODrNGqRYhqesOMRFK1S1WUdJw+dgxcObtJ7fi4rqscRzb7KxHjlORcxJNsrXKU95hdqAkxZxC8VSoIuRBOCNZgjPWXpCykJ5gKmmpx5Yja5yDL5FSLXxyvLS1OqY7m/T6mHuwXUW56xJSLLizFtp3nraslWREzEy66X+zolznyqxPKzLHuJC+5ExXUrkfBVc9WUKWbVtqUto4wW6PmLaJfw/YhMj7VOh+duhVtsO7gUKhiIMqQh4EX+pUt7MQCMwECrrOscpKO3c5rpS9sTKCFi+h313U3WKycgnBvFfFWih4Y4REuOIIWUmE7i9bsOd5lLKreVGu68nUfYSsY3wUValZVjKuPZg1ZxbXYH9SgdVnU3zGQPGIW0fINQ+PXFkYpVqEnGu9k3ZcpViEXHps5v9u9hqnUNwFH1c3gCIffMGo7jZ+Crfd+tveHvcQrpeN4pIlCJyU0HoxjkKOF7WWFSMk8rh86bOZZSS6zLHrE05DLQXm/lLvLUkxQmZFufZlWom0nAuucihHjP+LKtUYtfwIOrQKMm0TUojv+vwoAGDj7KGY0DOct1zczVJkFtdixuC2vGXkVlxclTWOExvaomTBmT0x4bp2uBPumPRIryeYvekCWjfzwxePDXB1cwRR2vud4nwkWYSWLVuGoUOHIiQkBOHh4XjooYeQmppqdZ+jR49CpVJZfFJSUuxqOMUS5svF3tk0Vw4OQm0XSobAJxjzHk/B7iTcWG+jq4QMOSyPTKuE2PrKahtTefNZA6XECDHh+tXe62tLZjyufYUQkyyBaRFjuiWK5VxGKbR6gvRbNaZtYk/pPE/2PSMzvj2Dt367jPisMt4yctzq7IkXV1mEuMYxW+qxtwZHIs3VzdmTeUoc+t3v3QRcK6rG8bRb2BGf6zZtpjRtJClCx44dw9y5c3H27FkcPHgQWq0WkydPRk1NjeC+qampyM/PN326detmc6Mp3PANOu42GAnFCAmejUQlUO5ZN3uVSMFkCTZYhJyBHAoYWxESt8+HOxNN//P1Ja9rnE0xQuLaxYdArgGr2JoxkBDCeT2Z23QiY4SY97c995q1fZk/ZZXWiionBN9zKTl0xQFwHdbe+0xUohgXxXMJxd6pLMq77h0mdGRHXkJ3tAjZMoZTKK5Ekmvcvn37WN83btyI8PBwxMXFYdy4cVb3DQ8PR/PmzSU30B1xWeYh1v/cLkL24mxDkaBFiOt3nv95jyFwPKUhRRF0JnLMpttiEbpZ0igo82dOtFER4lLE3cQ1ToxAIuQaZwtirVzb43Lg7+ONefd0t/1gNjaZEGJqpxLWt5EtRkjiubgqRoh7TSvz8s7tDVuVQke2UtEGPhFo9Xp4e3m7uhkUilXsSpZQUVEBAGjZsqVg2YEDByIqKgoTJ05ETEyM1bL19fWorKxkfSjCsBZbZAya9loInD0bJ5wMgVvY5dxfojTA9fKVY6FQKQg1WakWIeZhK2o1Nt03UhUhi2QHPJIDv2uchMZJaJc1dDbEQTUe2/rvTGFu56U8zmOy2sK4ALYmS7DYR+S24uoGrDx8DRetuL5xtUHINVYMYhJsuBpb2iImQYirELRmWimvrDNxHkTg3aR0hIzMFXUa5JbXOacxFAoPNitChBDMmzcPY8aMQd++fXnLRUVFYe3atdi+fTt27NiBHj16YOLEiTh+/DjvPsuWLUNYWJjp065dO1ub2aTgTZYgo4uFM1wp2LKs9ZlDYaVB+OSVYBFatveq6LJCL0RXxwidTi9G/yUH8OHOJMl1MFM4C2V0O5BUgM7/2cPaxteXNmeNE3AnYyJWaZCW8IBd1lbhlm+/RX8lm/4XGyMkNDEgdYworW4QOB4/klzjGP+zUqvzbHcmNrkAc9bD+F9hcjM7nb3l79YWynbEuZjfx9bcSl2FUpT0ugYddl7KRQUjHpMP5vMvlJJ/4JIDGL38CIqq1PY20WacPdFJUR42K0KvvPIKrly5gm3btlkt16NHD7zwwgsYNGgQRo4cidWrV2PatGn4/PPPefdZsGABKioqTJ/s7Gxbm+mxcAbX8g3k9h7Lzv0lH8/KC9HwO3fZxn1sd3URjhkRrtGWYfW7YzdE769Ui5BRQfjfgTQAwJazNyXXse1847MupNC9uCXOYhufksx3SWxZoNR+CyujLomKmC0JDQC2tWdSL+5MbVISMRjhuhZSBQuuPmAKU9YUK1t7Qs8zhggtFuswBBQD0dUIKBuS67O/isa6BCYmLJQfh7XEPXGlIvThzkS8/vMlvLglVtJ+1sYUZtzi1fwqe5pHodiFTYrQq6++il27diEmJgZt2/KnNuVjxIgRuHbtGu/v/v7+CA0NZX0obLhnqrlfgvbOaDljRqxBq8eh5EJUqjWCAqyQoiN1Fk2wPCs4XLg+R5vNpLqYOAvj/efjZdv5n71Rgt/jcizqk4Lx3jh1vRhphY0vV94YIRssQlIucEl1PRq0bOGaZfUyq6tKzZ5xNT/UlBXHUV7Lb0ERo/B99fhAzjK2xAhxHU9y3JPAca0pVraOTXz3ljG1t7Ox8zbjqVOaNdzRSHFpNnx3bOvcIW2yUtwDf7s9Lp/LsJ7pEWC32dqYwlSS/LzpkpYU1yEpWQIhBK+++ir++OMPHD16FJ06dbLpoPHx8YiKirJpX4oBoRen0EvH1mM56uXx1eE0rIpJx4B2zfH6xMaMglIsX6bfWWXFCANSXZVc+wYVaq/rFoW077jXi6rN6gPUGh3yyuvQuU2wqDr2JxXgq0PXEHuzjLVdTLIAI8x7nOta6wkxBdun36pGXYMO93990qJcZnEN7v7fUYzu2hpbnhveuL+eXReThNwKjOrS2vS9oNLSZeSn81l4+a6u3CfEA1PZ8PHmvn+5ssYJWVyZiSqMSB0juMpbSyhhq8WX6brIP2kEaHV6+ChBMLN33FaaEUXQImTNeuCIBsmDI8dbd0yfzexba5McOivPOIXiTCQpQnPnzsVPP/2EnTt3IiQkBAUFBQCAsLAwBAYGAjC4teXm5mLz5s0AgBUrVqBjx47o06cPGhoasHXrVmzfvh3bt2+X+VQofJYNOV157GHBjiu4VdWA758ebBFPseNiLgDgUnY5a7vwgpdc26QJSkIWIWZLnfEqEjqGoGvc7YuWWSyc1p4QgoziGny8+yrmTuiKwR1aWJTJKqnFkZRCwbr0euDGrWrkV1j3984TGRz7W1w2lu4xxE5tmD0EE3qEC8bhbD2bxbl9+d4UzBrV0fRdrdHhv3uu4khKkdX6uK71n5fykFtehxWPD8TE/x3j3fdAcgH0BDhxrZi1nS2EW+/Mt3+9bLHtelE1vjiQiufGdEZYkC9iM0uRnF+Jp0Z04K1Hw3Cp8/HiFvLFWoSY7nn5FZZ9yWXBqW3QARAvzDEFKPPabI2b4MuiZz4+1jToEBboXEVIyPJtTz1KgfD8b9pmpe2uPC2XHlshFiEpMJ8za2MKc97FlXoQVcIokhShb7/9FgBw1113sbZv3LgRs2fPBgDk5+cjK6tRGGloaMDbb7+N3NxcBAYGok+fPti9ezemTp1qX8ubONyWEu5Bx36LkP1DsF5PTPEfKQVV6BXFdndkrkZvPuOr0xN487pbcc0sMv4X4xrHbKdgjJBwfXKh1ugsFENzuPrc+O5huvlwtXvHxRws2pWESrUWABCTWoSMZdNwNLUIh68W4b1pvRDg640pK46jTqMTbG9JTT3uNlMMymoa8N6fCaxto5YfYX0vr21A8yA/i55MK2y0ED27KRahAT746vGBmNAzHAeThRUzJubt//74DWw+IxzDZFgstBqdWzdjbb+QWYafzvHvTwhB9e3ravxuJI6RJc1cTmAqEWmFVThzo8SibuOkwYHkQux7Yxz+seYMAKBti0De9hhjX1QqgO9RWn8yA9HN2XVwKZ5G4eZISiF2Xc6z+N14CjX1jeefW16Hrw9fQ00D131keYxqxr6f7U9FVPNA6PQE0c0D2PE9kq25BvSEoEqtwZu/XGIpiYD1WWxjXAP/eASsPZ6O/+5Jwfn/TER4aIC4tnGmaW885rvbr6BLm2C8NL6L6HqUpBTlltfhj/jcxg2iXOP4fxMiNrMUUc0DcUdz/mdCKag1OgT4cqeYlvouE6JSrcGlrHKM7tra6j1sD0zlx1qMENMiZEt8oqux1m8U90Kya5wQmzZtYn1/55138M4770hqlDvCXJfCKccT2CY0+ybpWDKMURrG9E8thzDEnKRmHq+iToORyw5jfPc2+OzR/ha/C8VwmAsYRVVq5Jer0b9dc846BK0xIq6mvXeBcf8fTmdi2d4UyzYInD+XMsf10ptnZm0w7jZ74wUAQPMgX7w1uYcoJQgwKAfmfHYgFXsSCqzul5RXidFdW1stAwCVai2e2XQB70/rhY93i8+yx0VKobjg3P6LDwCAheIOAKti0nn3W7AjAT9faEz8YEwgAbD77JuY66z9mEPI5C/5M2sChgkFJmmF1bz35/nbvv3eKhXvOJVTVoeXOBJQcKHR6fHsJu7A6d1X8lFUedoiNup/B9M4y8/ZannMQ1cbLXU3imvw0KpTAICIUH+8PblHY0HG6Qq5tLEUKD3w7fF01nGMcLlLanV6qFQqvPP7FRxJKcSheePRKtif8zj/3WN4ZqeuPIHY9+/hbQ8TazEzKQVV+DXWEKPx/NjOVgVYYmXs46K4ut70f4NWDz8f/utnnLCQil5PMNps8sPYtnqtDv87kIax3VqzrNEqiGt/Tb0Wh1OKMLFnOJr5G8SZ5LxK0+RA5vJpktvrTLadz8KCHQn48rH+eHigZbw1W+ay/0U8c+1ZJOVV4r2pvfDCuM5218eF3oqCs+7EDdRr9Zg7oSur3KbTmRgl4h2gFC5kluKJ78/itbu74VWGKz/FPVGAI7T7cz6jFIM/PoSdl3KFC8uEULApcwBNzqvknrm141hSYc661ms5FCGmRYhxvF2XclFUVW8K1gSEYwS4ZkW1Oj1yy+swbOlhTF91Cgk5FTzlDf+rNTqcu1ECrVkWKVuuBSEEf13Os4iB4SP9Vg12XMzhVIIs2iuyDWJm/1qbCXZfH7nOU1I8+SLc4KS6btqrBAGGdLDmlNY04IGvT3Jaiq7mS1vLjKkEAZYKDx9/xuciOa9SlFsjYFiU1EiDVs+bWOCNXy4BALxkmAUOC/RFjIBL4YXMMlxmPGNyUVhZb2bBNfxdsOMKBn500GoaXvM1qoqr6jnLmVuE9HqC+78+ifu+Oo7tF3NQVqvB0dRbgm29o0WQYBkjXPURGBTO2oZG65hQVju2SzR3mep6rUlJZZb/NdYyOyvzbhmw5CDiBdZ84kLDGXtm+LsvsQBrj9/AU+vP49yNUs4yAL9S9OXBNLy2LR5zf7po2paYJ+9950jL2oIdBmv5m79YusAC8luEkvIM49jms5lYtucqEnOlX6tKtfUU2kzXWaZ1SK3R4ePdV/HZ/lQUValZz9kVB4wVQmyPy8G4T2MEXbm5+Hj3VWh0hHdyxxqHrxbij/gc4YIUp0EVIRl4dtMFlNY04PWfL7m0HUaBUq8nJncnANhwKgOvbYvH6evFfLtawBSomC8hW3Lud5y/G30X7jd95zKDezMUod/iGl/IzBnelYevQa3RWbwQNDo9v1vg7b+zNp5nzUqeyzC4HJXWNOBXhtBqLP/u9it4bO1ZrORQBgghePO2YMncZoQ54b7y8DWM+SQGr26Lx6Qv+ONJzDG31jApr9Xg2U0X8PeVPE4lgmublwhrJSEEJdVs4fALGwZ6Jol5wgrEvsQCvPP7ZdSLtDzZyvztV0z9xBQujQz66CAScitc6qbx84VsTF15QnT2srd+a7xPGrR6QX93bxms1hV1Gs7U5c5iNUOpNI5N285no0qtxd2fH8OLm2NZky1FVerbLm2N/Xr82i3W5AoTc4tQpVqDlIIqlqumMbvfjos5WHeiMfV9BkOBbRnky3sOTNc/wDBGm/NNzHX0/GAf3t3e6FoqJbPfqevFWPJXMuta1DZoMWDxAYz99IhFea54L/PDrTtp2U4huJaSMVbLTLbxzKYL7DIiXOM2nc4EwFYk/a1YtZg05dCQ7NI6fHf8BmeSF3MOX2W7Id+56AByyiyTpBjhswgxlfgqtdbl7nBv/XYZWaX852FOYm4F5v16CXnldSz3YqlJLJ77IRZv/nJZcCFZoWya9qLR6bE3IR9PrT+HWzyTQuaQ24mCPA2qCMmAWNchObG2AN/zm2Px2rZ4i99TRboDWdRrx32fwTGzffJaMX4zm31kymf7kxoHXqYV5YuDafj2aDrrzKvVWgz/72G8sLlRMOOyjJ26bhlrARhevmWMReKMwvLOSwYL2tdHrrFciT76Oxm9P9zP9ncHcOfiAyitacCtqnpWXM8XB9NkXzn76yPXcCSlCK/8FM/ZN3nlaovry7QI1dRrMWvDeYv9SmoaMPjjQ6xtKw/zp7kXg5gB9sdzWfg1NgcHJMb9SOXnC9n46G+DNalYYBFPd8QwIWC9jNS4gIo64QUUnU0mQ3heefganmUI0NX1WhxILsS2c4Y41T0J+Ri29DBmb7zAEsSsTVqZyx/W5JF5v17Gx7uvIv2WYZxiWkz4dvv++A30Xbgfu6/k81cM4HJ2OXR6whoDtTo9/ozPxXObLlikWjc/5mf7U7HhVAZWMSZzrhVWQ6snKKy0TOteU2/5HjMXxsRMqJjDtaim8T7lux/Nrx3ffc0Vo2HNRb2mXovT14stLP3WcGUwvRKSJTz3g6ULrPH9yAVTWWf2PfN+m/i/YyYlFhD2Cjh7o8T0jDkbvd4g/N//9UnsuJiLd7dfYf3O5RJurS4jZTX876DkvEoM/OggNghMPOj1BIm5FZLXP8str0P/xQfw7x8v4sS1Ysw3Oyc+Zm28gAe+OelyJVZuqCIkA8yb4ufzWRazfWL4dB+3GxRguNmvFVaZhHpCCJI5ZtqNv/NlwioTsSo0F/bc8lyzB98dv4H/+/0KUgoaz0GsgJaQW4F0hmBwILkQpTUNOMSYtZISZHvZLBnBhcwyllsTIexz+PFcFqfiW6XWYlXMddz9+VGrx5RjMGcOvFmllormptOZ+L/f2QMbU4DZdTkPx9KEXXucjVT3M1vYcCoDOy/linZTdCcadHrBuApHBUi7im3nsznHu1u3LZsv/2hwmTqWdku0NcVc+OcS5AGwhGmjwm/uEsmFMRPiG79YTlYJEZ9Vjjd+uYTDKUVY8ley6T4+lFyI6atO4ZcLllkTrzJiyQL9GhUHc/fQeq0earOxzdw6ptboMP2bk/hkXwoq6jTIFjGjziU0GQVfawsEs10guctxKSnms/XM/pyzNQ5PrDuHjacyrbZ569mbjPet1aIOhSngSm2H2Jn7D3cm4qFVpywUY1thXm/mo2OelITpeu3vyy+KZhTX4PG1Z61m6JQCn/zEhU5P8OCqk3hy/TlWe5iXVqw1BWA/T9a65z9/JKCiToMlfydbrW/N8XTc//VJvP0bvwcJF98dS2fFah82G0NrG7Q4e6OE7U6sJziedguJuZVILfCsBXCpImQnX5vNmM/fkYB6GwaU1Uf5A69XxVzHPV8ex9rjBheM/UmFeHj1aYtyQuNex1bifdbZ9dr+JrAme+SWNVpKxM40HkkpYrkDMV+X2aW1BtMt8xjldVgv0Z1jqVkMyt8CM7dGcspqUSWgBE/83zGbFGU+Jn1hPaDeCDOGnGsmWQlUqeW7LtZwtQuro9h4KlNQwPM0RYiPVTHp6Dh/N2sb15pHXJgL7lzCuiF+hxn3aBjzzzMWnDyaeov1frhxqxpTvzph+m4uGIqB6T72W1wOJn1xDPeuOI7nN8ficnY55+w0U2Fj9r55rOa281kY9NFBZBbXYPeVfKg1OotrcTztFi7nVODbo+nov/gAxn4ag3d/v8LpVmeES1CsqdficnY570z2oauFrMkaY7lP9qXg1W3xyCyuwWvb4jnHDKb79ujlR/DsD43XzJjKfuu5m1anDN7/M1FUHJg5JdX1rOv66b4UfLgzUXI9Ru5j3C9SMiSeTi/G0KWHsC9R+N21+cxNXMouR0yq9bg/JlyuxUbEWITMeXxoe97fMoobJ63skUVqG7SIu1mG32JzMHXlCeEdYFB6EnMrWR4lzfzYOcb47mFCiOXEgtjJGBEWsj/ic7D6dsIeaxY6LoTGnmc2XsDja8/ie4bbL1OJa5BogVI6krLGUdhU12ttCpaTivEYy/am4KXxXSzcnowIjRHNGT7ry/ZcxbWianz/9BCcu1ECf18vDO7QsrEujoD8y9nlSJAYXGntgfbz8UJtg5aVuEAqzOrHfhqD1yd2Q5/oxgxf7//J/RIihN8Ht0dkCOJuWgoUQjBd+qyRU1aLnpGhaBHka7OVTireXirklNXih9OZnFn7KE0DW1ybmhrM52P3lXxWIL6RbeezWFZFLqUBMIzd47q3Qf92zfHO71eQ7ACrp3n2QHO0t1179icVYF9iYwZHrgm72gadKT5t5rB2FufUspmfRXD5L7HZiMsqw6F54zmPzxXvyJcIxsiVnApWAL1WT/DrhWx8e3vC8O8rebzvO+YtnlehRl6F2iKjoFZHWPtP+uIYcsrYytwzmy4g5aN7rbaTSUGFGiOWHcYdzQNxav7d0Oj0pgnOF8d1RpsQf2SX1qFrOPfi0MyMalwQYnBVza+oQ89IyyyWTJ5cdw56AszZelF05jwp7k6rYtLxf1N6CtbD/L9Bx//e8bEyQcMcs3R6wrsgtBCzNpyX5MZWpdbgR44lEswTzvApQp0W7AEAHJo33tTnTCXRXLFNyquAWqPH4A4tBBWhx9eeFT4BGM7Bx8uLZQkGuJVSY+bjoio1zt2e0Nl2PgtzbqfsZ/crVYQot9GJmNH7Mz4XPSJDONPv2grfQyI0Y8S8+b+7bV06mFxoSl97fel98PZS4bWfL+EvRpY5o3vY9NspbE3t0BPBLFTWXB8yi2uw7kSGXW5a5rV/JTKmZemeqxjb3TXpOo1jSERogNMUIS+VCi9tiTNlDaI0Taxkl6bcZurKEzjw5jh0jwjhVIIAQ6pyZvIEa2nHp686hQHtmnOuCfbu71dMaZ8dxYlrxRjy8SGUmMUkmM9Wm7PtfLaF6xmfIs3lalql1iAkwFfy5BkXp64XY8WhxrGd6xWYlFcBfx8vznWylu9Nwfv39zZ9r9fqWNYFPlfZH05nssrFpBRhXPc2LMuq8dfj1wzvsdzyOjRo9SxriEZHMGvDeZy9UYp1Tw/BpN4RFscyZsN8dEhbhIdYrj91s6QGYz9JQKVai51zR7OWgAAMlj+NjiDQz1twIXIu6hp0uF5UzauomXPqejEiQgNM5Yur67FoVxLCAhsnXN/dfgU/vTACEaEBVj1lrMW4XMwqN/1fVqtBSkElRnRuBV+zwUyr06NOo0NIAHeSErFK0KnrxUjOq8TFrDLsTbRc+sH8/jIqCBW1GoQF+UKr07Puj0lfHDMpo0yZ0TyWedpKQ+KKSx9yp9w/lnYLV/Mr8ZLItOdqjQ4DlhxEMz9vXF44mRU7t/2iZaKYBp0e/j7e+IrxnOlvJ0f44XQmOrRqxtguqgluA1WE7EBM2t83frkEX28Vri21bQHZ//zBXozyWmEVYnhM9kI355ytF5H+36msh5S5hkelWguNTs9SggAAhDshRP8lB/Do4HbYeu4mvp45EFP6RHK0ib9RH+xMst5gB8N0Y2EiJTOTPdRYcS+QG28vFVWCKLJkjWsKfHkwDd8+OVi2+vgWRv6Fx7ovN+ZKEGBwrxHCcpFT/rGxwmxSp9+iA+IaJwIxMQlGQXLNk4Msflt3MoOlCBVXNwi6MQOGtbWYGVif2XQBUWEBOLNgokVZ5qRf9/f34pcXR5i+qwCcvdE4y86lCBm5eLMcgzu0QJsQ9nIGnzPWIjuaeoulCN0sqcEzGy8gs6QGVxZN4axXyK3M6HL+M6Pd1vjXOkPcjFEpW7r7qoUbefqtGry0JQ5/zh1t1TXO3FWrtkELrd6gPMYzFKFXfrqIcxml+GRGPzxm5k7X84N90OoJ5t3THa/ZuLbPGz/H408BNzMvlYp1LTV6gu+P38DSPVcxqVcEEnLLMaxTK859WTFCjO1Md/mSmgaL5+6rQ9fw5SFD/zM9Xsyprtfiwz8TcU/vCHSLCIbudgbh2gYdmvn7IDmvEu9s544nqtcaFCGmwkoIEJ9djkV/sWOVaLIECirqNKio02CVyLVBzB/ya4VVgnErJdX1+M8fCfjpHDv49R4riyzuScjHMxsts4Ex+T0uG/d/ze0f26DVI4ljDQYCwumLXaXWYsOpDDRo9bwzoo5WKqRmS2HCtZYMYJlAQW5WxVxHx/m7kV0qbzY5a1CXKAogzzpCTYG9iQWY8a1lHKYnYctEVJ6VNVdmCbx77EHKWi9875zp37BTRR8UkaVyy1lL16j8CjWnF4N53MxjDPclVkppQlDXoONVDOZsjcPQpYc4M78aYQ7nf8bnYvxnR3GjuAZ6AlxiKA5GymsbMMpsUVs+fouVtsbN9FWnDAuV88SJGScBrFmEbhTXsJJBTVt5EncuOsBSggCYXLa2x1mu2Wjsd3uWfBBSggBDwibmGmland6UAOXQ1UIUVtZbTibDYKFhutwb74m88jp8ui/VtL2ost4i06xRCQIMa6nx8dWhNOyIz8W/f2RbsstqG5BZXIP3/0xAYi73hOiY5UdwvaiKFUdICDgnUN/5XVyWOXeBWoQkUlSpxthPY2xKiGDEmjJj5Invz0lOd/3Z/lTBMsw1KczZnZCPYH/LdKR6Ii6N8opDaUjKq8SskR0xppvB7czRMwfG4FdbkLKGgJzsThCXfEFOzNN9exprnhyE/+5JcVmfugtUHxaPLXGCTRk+q5ez667XcL+b5Vzgd8UhS2HbfEFqJj8yJjSPpt5Crw/3oV3LQKvHsLYIuvEx/jU220IoNU+IQgjBgaRC0cpkea30pQUeWX3a6vkD1hWhvy7nYXinlnhyRAeoNXrOZTeYGL0pbtyqRtsWQfDjWDtqy9mbaBPsj3t6R5jWDXQE1tz/jdwsqcH4z45y7vf42rOs99bM79nxP+ZJX6zFU+WVN/Zxg7axXW//dtlkkeSjUq3F+38moh1jIejc8jp8wBFnLWfCJyVAFSGJnEovtksJssbrP8fjoQF3wNtLZfOaP/bw0d/J+GRGP4vtOy7msAZyPow+3AeTCxt9YhVsQhVzThT3wMfLy2VKUOtgPwDusTZRkZXZRArFE9jMYcWRG6alQq3R4dcL2XjHylosXB4g9ngExKQW8SZqMhekh3x8CEM7tuQsy4V5KmUx5JTVIbo5t2JnnHwRSs/9/p+JeHJEB1FpvG/cqsHhq4V47odYTO4dgZbN/Fi/p9+q5hTgHYFGROIAcyUIaMyoJ/W9xTeZtXBnIktBYXr3CClBRtQa4UW5PRGqCEnE1lz7+xIL8HtcDl69mzsjDGBIgSg1DaLccFmMxLoAmhOfVYYz6Y6biaFQjNiaSUgOVv9rMAa2b45u7+11WRvE4qhJHApFKTjatdmcwsp6q0qQI7jI4f7GR0lNA/YlWQb9yw1fzK3RLds8XbtlOcPfeivZ5YzUaXRYc8yQkY9rIW5nLgTNdGuTglZHcDFLutWZb/mHH86wJwDM1xIUA5dlrSlAFSGJ2LrWiTEpAXPhT3fBlsxmi/9KElzThEKRC/MMQs7E38cLvt5euKtHG5vWHqFQKBRPxThFxeeyaMSY3lzMZHPHVkFWy1lzH1MKT29wXEydrVwvquZVaD2Zpqn+2YGt2n9To6koQX40H7HD6dTakLYzxN8HiYun4N7b2QkDfRvj2Xy8VPjsH3e6pH3GWbSNs4fijUm2ZStyNl/PHGjVOu3uBJmtm3F3z3B89fgAzrI9I0OQsGiyE1olDrHrvlAo7oDJNU4gsZHvbeXl1HXhuN/MklqrMV//WHNGfAMpJko5sktyYc/CtkqESnES8bQVdSn2EcSRXIJiScayqTYpjXPGd8GBN8chcfEUxH1wD4L9ffDJP+7EvHu6Y8/rY03lfLxVeHRIOyQsmozrS+/DiXcm4P+m9EBUmOV6HPZw9O27LLb531aEVCoVaw0NV9EiiL8NY28nMXmgfzTemtzDWU2SnY2zhyIilD84O/b9SazvG2YPxfQBd+DQvHEWZQe2b4GQAF9TrBeFQpEPHy/D+FgvsG6V0SJkLaGTWGwNYaA0TagiRKHYQSuzIM2LH3AvhtbUUalUmDG4reT9Xp/YDb7eXgj29zFZXsICffHaxG4mSxEA+PsYFNKQAF/4eHuhXcsgzJ3QFafn321Rp1E5mndPd8nt6di6GQ68yRam/RmWKWuLY0q5N+7gCTwWYunDfXHhvUksaxlgsFr99coYfPeUtLVxHh54B/q3DbOpLY6kql5rNY1skJ8PVs4ciM6tm7GUoq7hIab/778zCvf2icTiB/sAkFd48rUxZu0XkWu4UCjuQp1Gh7ibpYKZ/2x9ZijOR+VhGRWoIiSBokrxaxlQmgYtgvzw+NB2pu/m2WvkYu6ELoi2wbrx7r09ZWtDxrKpOPjmOGx8Zihruy0KhRD/N6UHEhdPQaCfdYvbzGHtMKFHG/SO4l5kjmvAnjGoLVI+uhevTezGaSEQontECH58frjpuz8jwNTcJYtJy2Z++PihvoL1PzLwDux8ZTTv7y+M7cS5vX/bMDw2pB18vL1AwHZdIISgX9swBPlJCwvVEyLYB46iVTM/PNg/mvO3EZ25s2B99fgAnFlgUH4f7B+NI2/fxZvWd1inlljz1GCTgm2+3ps97H1d+n0FAMM7Wy7E+PTIDhjAWDxTbk6+O8FhdZvTKyoUvXieVYrnMuPbM4Jr9BRXN1ikiqYokzsVODlmD1QRkkCbEH9sfnaYq5shyKIHegsXcgOsub6kfnyvE1vCT9sWgfjvw/3QMzIEz4zuKGnfnXNH472pvUSVnXdPD5xeMBFPj+xg2vbIwDus7pP+36myzrKpVCp0iwhBuxZsa4XYRAV6CanUg/y8EWzFumJk2SN3YuMzwyQtFPrQwGgE3LaYdGkTLFh+YPvmFtuYC9SKVYQAiBJo//fP/mgd7I+2LbitQu9N436+d74yxuReYn6pbRXytTpiYV2Sk06tm6HvHdyC8YSe4Vg5c6DF9W8d7IfwEO5JgekD7kBUmDhrGnO9DEBet+f2LYOEC4lkQo9wUc8CF13Dg/H5o/2xcuZAPDeGW4EWWv9FTkICfCSNAxQKRXl8NF14Qs+doIqQBFQqFcZ1b8P5W+c2zTi3u4LZozthzvguostzxTXMHNaOo6RzaRHEb10xukLJycSe4ZL36RUVCi8vFfa9MQ4LH+gjad/gAB9RAnzGsqmmRfKeHd0ozPxjSFuLxfOYeHupHBLT5ufNvvZ6kYGTUhZhc8SM07BOLXHhvUks9yiVSoUv/tnfouz0AdGmmKZX7+6Ksd1a4/1pjUorM/aIeS8GCCgNfaKFZ8ONVqwdL4/C3Tz35EvjOlutwzyYtb+NFoUOrYIcKij7eXshxN9y/Pnw/t5YdNtljal0dg0PxpbnDNa4kRzWEzFsemYoFj/YB+PNxnJjEg45kHMCok90qNXn3BpT+0biH4Pb4sH+0Zg9qqNpO1PJ9rcxZW7PyBDhQmZ4q1R00WMKxc1pJ+NEjxKgipBMvHVPY+Dx4bfGy2Kx8PfxgkoFzBrZAYM4ZqW5MPrzvzW5O358fjgm9TIIUsM6tkQzP288MsjSihDz9l0sBW9M19b478P9cOlDy5gGPhckuYkKC0C/O6wLwyfemYCDb47DvjcMQfPT7owy/bb936NMbf3q8QFoE8ItzH1wf+PsujVhwzwd54vjOmPmsPZ4ZrTlLOuJdybgjUndMLJzK97ZbgAI9veB2koA6UvjO+PnF0ew3Ls6tm6Gswsm4sfnh2NUl9b436P90SLIF+tnDcGc8V1MsSXGWXS1QMpSW2C6SkWGBqC5leB8wKCAAEAPEYLTQwOise7pIRjcQfwCgGIJ9vfhvA8e7B+Nqf0MQnBEqD9SProXXz0+EN88MRDvTe2FCT3CseW54Xh+bKPy0bF1M7x8Vxe8Makba+2FvneEoXWwv8kiMO3OKPSJDsWKxwYAMCg5k3tHmMr3jgq9rQxYzviHhwRg/n3cro3z7+uJo2/fZRoXzGN/mJPurYP98eJY64qTeawbACx+sA/eubcn3pHoXvnSePaxjPdHiyBfhASwz3P+1J54coTByjljUFvMndAFO14ehWfHdDJZQTq0anzpHpo33uRatebJwVj1xCB8OsOQLVCs+9hdPcIxa1RHi0mIpQ/3xZLpfWRxb1WpVOgazm9tHNyhhdX9/5xrcI389l+DEB4awBp/3p/Wy2Qtf2lcZ2z/9yje+u7u1XivMRXaB/pH49nRnbDthRGi/P1X/2sQgMb75O3J3fHrnJGSFTRvLxWeHtWB9/eFD/TGzrn8bqHmcFlr5UCqdV8qXSROnP4p4Zq4G0pIMEMRz+ePWk4cujsqYkMevNWrV+Ozzz5Dfn4++vTpgxUrVmDs2LG85Y8dO4Z58+YhKSkJ0dHReOeddzBnzhzRx6usrERYWBgqKioQGup6/2JzP9bRXVthy7PDMX/HFXRo1QxzJxjS0h5IKsD6kxmo1+p5AwXbtQzkXGF6w+wh6BEZitLqBvRrG4bqei1GLTuMdi2DkJRXaSoXHuKPaXdGYWq/KMTdLMPUvlFozxActDo9ruRWoH/b5gAMi5Z1f3+vyVXmgf7R+HrmQOj1BFcLKtEzkj37uOJQGlYcugYAmNCjDZZM7wtvLxXis8ox96eLAAwvjZAAX6w8fE3U9et3RxgCfL1ws6QWT4/sgD7RYSiqUt8WHFTYeCoD/5naC95eKny+PxUdWzfDZ/tTERUWgPwKNeaM72IhIGp0evh4qdBpwR4AQPwH96CFmUCz5K9kbDiVgfem9sKJ68V4ZUJXDOvU0tSfs0Z2gJ+PF74/YVgFvFt4MIqr67H/zXEIDwnA98dv4Osj17Dp2WEY1N66IGOkoEKN3+Oy8c+h7TBs6WHWb4mLp2DN0XR8w7Fg7V+vjEE/G6wi2aW1+C02G8+N6YywIF+88tNF/H0lHwBweeFk9F98QHKd7VoGYvucUQgPbbSCLN2djJ/OZeHU/LsR5OeDd36/jHHd22DDqQwk5laiY6sgbHtxBLJL69CpdTO0CfGHRqfHD6czMaJzK8TdLMPorq0w6YvjAAzZzE5cK8aZBXeLdm0Sy4/nbmLt8Rv44Zlh6NiaXwDR6wl0hNi9JlFNvRZ+t9cW4oMQgvJaDcICfVFRp0FIgA92xOdiZOdWFrNtu6/kw9/HC1vP3cQDd0azkk7UNmiRUVyDPtHse2X+9iv4+UI2Zg5rj/8+3JdX2L2QWYojKUWY0CMc//yuMeXs1ueGY8ztDHPG4xRXNWDcZzGc9bQJ8cetqnpM6hWBdbOG4IGvTyK1sApnF0xE80BfnM0oQe+oUIQF+uJo2i08s/ECIkMDcPY/E0EIQXZpHdq1DORsZ3xWGRbuSsKzozvhIR530MziGkQ1D5DFUrwvMR9v/XoZAb7eKBGZTtaiPcun4WZJDVYevo454zsjunkg+izcD8Ag+B3/vwlILazC+pM3ML57OL47no4Ns4fyumou3JloWjAxc/k0FFaqcS6jFPf3i4KXlwo5ZbVYticFd/cMx97EArw1uTt0eoK+ZpNJn+xLwa5Lefhz7mjWpMBj353B+cxStA429GPM23chMjQAn+xLQatmfnh1YjdUqTUI9vdBYWU9Im9bRM9nlMLbC/hsfyrO3ihFoK83ekSG4FJ2ObxUjQp5x1ZByCypxeIH+2DWqI6IzSzFnK1xKK5uvL6/zRmJoR0NEyBanR5aPUFaYRVWHLqGaf2ikFNWhyeGt0frYD+oVCoQQqBSqfD8D7G86/M90D8af13mjk2ZPaojqtRaDOrQHGuOpZvew+1bBuGvV8Zg1+VctA72x79/vAgfLxVmDmuPLWcbF6308VLhg/t7Y+GuJNO2PtGhaB7ki16RoXh2TCfE3izDa9viWcddcF9PDOvUEot2JeG9ab2x5lg6quu1yC6tRUGlGkaJrHtEMNIKq9GlTTMcfusu3PfVCVzNr4S7Y/RWMS6IuvapwXhxS5wrmySaD+/vjSV/J7u6GS4l9v1JTnWntYZcuoFkReiXX37BU089hdWrV2P06NH47rvvsG7dOiQnJ6N9+/YW5TMyMtC3b1+88MILeOmll3Dq1Cm8/PLL2LZtG2bMmCHqmEpThE6nF+P74zcwuU8kdl/Jx5eP8VscjDRo9ej+vmHl+fn39UT/ts0R4OuFge1boKJOg6lfnTD91j0ihHP2vF6rg4+XF6rVWugJwcnrxbind4SgO445RqEmPrsM9/WNsrqasE5P8MPpTAzt2NJCMF8Vcx17E/Px4/MjEBrgg6zSWsRmluGt3y7z1vfn3NGmmVudnkieUcwpq0VkaIApFsKc/Io61DXo0JlDoKjX6pBfrrYQhn+Py8Ef8TlY/cRghN2evc4rr0NUWICFYKbXE0nxKEyu5JTj6yPXcTW/Er2iQrH2qcFIv1WDh1efwsD2LVBaU49HB7fDLIYLi70cS7uFWRvOY8agtvjfP/sju7QWN4pr0DU8GIm5FWjfMgiHrxbin0PbIb2oBl3Cm+FKdgWCA3yw8vA1rH16iKT4hIIKNX48dxNPDG8vSaEhhECt0bssMN/TqKnXYveVfEzoGS44Nhm5kFmKdi2CTEIuF/N+uYSCSjUq6jS4q0cbpORXoUGnxxf/HIDdV/Lw8MC2CAvyhV5PoNbqOJMzEEKQlFeJjq2b2Rz74mj0eoKs0lq8su0iCivrcauKnaFOpYJJYDUK/Hc0D0RueR3ubBuGXa+Msagzr7wOh68W4h+D20m+z2MzSzHz+7N4YWxnyRY6MdQ16JBXUScqZo6PmnqtKWsiIQQNOj1e+Skeo7q0wj29I3AxqxwP3BllGlNXHr6GLw6moWt4MA6+Oc7mTFSEEHy8+yp6R4XiZmktVh6+ht5RofjH4LZ4ZNAdGLDkIKu8UVlnotMTvPHLJfSJDrVwKy+raYCvjyFzZV55HUYtPwIA+PvVMeh7Rxh+Pp+F0toG9IkOw4B2zS0sHGfSSzDz+7N47e6ueGl8F6uZJQGgpLoedRod/Ly9sP5UBh4d3A5dw4ORVliFyV8eZ5xHOMZ2a4MjKUVIzq/E4gf7YN6vl0xeAJ1aN0NGcQ3vcd6c1B1fHkqz2pZZIztgSt9I6PXA4ZRCBPv74EpOBa4VViGvQm2anGzfMsjk8ti2RSDqtXqLZ8bIhfcmoZm/NwYsOYixXVtj/eyheG7TBRxOKbIoO2NQW2y/mIN7ekeAEGBUl1ZYvjfFwuW7Q6sgqADkltdZjYd8fkwnnEovQZsQf/SOCsU/h7SFRkew81IucsrqMKZra5TVNmBvYgEuZZdjUPvmqK7X4p9D2qGZvw9mDmsPQgi6vbcXWj1BsL8PBrZvjhZBfugRGYLjabfQOtgfuxPy0b9dc8wa2QFaHcHOy7kY260NTqeXoG90KHLK6hDdPBBrjqVjUq9wVNRp8K/hHXDoaiG6hgfjp3NZeKB/NE7dlvG2nc9GkJ83IkMD0Ds6FNeKqtAjIhRH04pwV/dwXM4px+Xscmj1BLNHdUTszVL0uyMMp9NL0DzID+W1DWjfMggnrhXjP1N7oqSmAdVqLX48l4VhnVrifEYp7mwbhis5FVj6cF+EBvji97gc1Gl00Oj08PX2QutgP0SFBbK8aFyNyxSh4cOHY9CgQfj2229N23r16oWHHnoIy5Ytsyj/7rvvYteuXbh69app25w5c3D58mWcOcO96FV9fT3q6xsfosrKSrRr104xipAjMM5ueQpFVWqEBvjCx0sFlUpls497U0Cr0/MqdnKQW16HiBB/hx6DQvFU6hp0yCmrRdsWQbhVVY9WwX4oqqpHy2Z+qKjVmCzwej1BpVqD5lZiG+1BrdFJnvRSMoQQFFbWIyLUX9Z3n7lCZvQS+Pihvmgd7I+7e4ZbnfwTghACnZ64ZDw1TsTxyQv1WoMCdb2oGu1bBcHfxxv5FXX44M9EjOnaGr2iQtG/XXPTfaTXExRVNVr4SmsacONWNfx8vNArKtRu67gROeWbSrUGwX4+qNPoEOjrbTExaT7BWq/VOSSm2NNkNndELkVI0nRcQ0MD4uLiMH/+fNb2yZMn4/Tp05z7nDlzBpMns1ftnjJlCtavXw+NRgNfX0v/0GXLlmHx4sVSmub2eNoDxZfViWKJo1+otq5JQ6FQDDFx3SIMFnqj0tPptqDNnP338lI5TAkChBNxuBsqlcqq9dFWmBYXlUqFUV1aIbWgCpP7RMjyXlKpVPBx0Zo3RqGfT14wCvzG+xUAosICsW7WUM7yXl7sPmjZzA8tm8kfnymnfBMaYHjm+Cxr5pOujlCCAM+T2ZoykiSw4uJi6HQ6REREsLZHRESgoKCAc5+CggLO8lqtFsXFxZz7LFiwABUVFaZPdna2lGZSKBQKhUKhYPOzw3D8nQl0co5CoXBik4O2uSYsZCLkKs+13Yi/vz/8/ZURjEWhUCgUCsU98fH2om7BFAqFF0mjQ+vWreHt7W1h/SkqKrKw+hiJjIzkLO/j44NWrWxbB4JCoVAoFAqFQqFQ7EGSRcjPzw+DBw/GwYMH8fDDD5u2Hzx4ENOnT+fcZ+TIkfjrr79Y2w4cOIAhQ4ZwxgdxYbQgVVa6f+pICoVCoVAoFAqFYjtGncCGVYDYEIn8/PPPxNfXl6xfv54kJyeTN954gzRr1oxkZmYSQgiZP38+eeqpp0zlb9y4QYKCgsibb75JkpOTyfr164mvry/5/fffRR8zOzubAKAf+qEf+qEf+qEf+qEf+qEf+iEASHZ2tlRVhoXkGKHHHnsMJSUlWLJkCfLz89G3b1/s2bMHHTp0AADk5+cjKyvLVL5Tp07Ys2cP3nzzTaxatQrR0dFYuXKl6DWEACA6OhrZ2dkICQlxeaYOYyrv7Oxsj03l3ZSh/evZ0P71LGh/eja0fz0b2r9NA0f1MyEEVVVViI6OtqseyesINXWUtrgrRV5o/3o2tH89C9qfng3tX8+G9m/TQOn9TFOpUCgUCoVCoVAolCYHVYQoFAqFQqFQKBRKk4MqQhLx9/fHwoUL6TpHHgrtX8+G9q9nQfvTs6H969nQ/m0aKL2faYwQhUKhUCgUCoVCaXJQixCFQqFQKBQKhUJpclBFiEKhUCgUCoVCoTQ5qCJEoVAoFAqFQqFQmhxUEaJQKBQKhUKhUChNDqoIUSgUCoVCoVAolCaHRyhCy5Ytw9ChQxESEoLw8HA89NBDSE1NZZUhhGDRokWIjo5GYGAg7rrrLiQlJZl+Ly0txauvvooePXogKCgI7du3x2uvvYaKigpWPUuXLsWoUaMQFBSE5s2bi25jQkICxo8fj8DAQNxxxx1YsmQJzBP2rVq1Cr169UJgYCB69OiBzZs3S78YHojS+1etVmP27Nno168ffHx88NBDD1mUyc/PxxNPPIEePXrAy8sLb7zxhtTL4LE4q38zMzPx3HPPoVOnTggMDESXLl2wcOFCNDQ0CLZR6Pml/duIJ/TnyZMnMXr0aLRq1QqBgYHo2bMnvvzySxmujvvjCf179OhRqFQqi09KSooMV8i98YT+nT17Nmf/9unTR4Yr5Bl4Qj8DMsnNxAOYMmUK2bhxI0lMTCSXLl0i06ZNI+3btyfV1dWmMsuXLychISFk+/btJCEhgTz22GMkKiqKVFZWEkIISUhIII888gjZtWsXuX79Ojl8+DDp1q0bmTFjButYH374Ifniiy/IvHnzSFhYmKj2VVRUkIiICPL444+ThIQEsn37dhISEkI+//xzU5nVq1eTkJAQ8vPPP5P09HSybds2EhwcTHbt2mX/BXJzlN6/1dXVZM6cOWTt2rVkypQpZPr06RZlMjIyyGuvvUZ++OEHMmDAAPL666/bejk8Dmf17969e8ns2bPJ/v37SXp6Otm5cycJDw8nb731ltX2iXl+af824gn9efHiRfLTTz+RxMREkpGRQbZs2UKCgoLId999J/PVcj88oX9jYmIIAJKamkry8/NNH61WK/PVcj88oX/Ly8tZ/ZqdnU1atmxJFi5cKO/FcmM8oZ/lkps9QhEyp6ioiAAgx44dI4QQotfrSWRkJFm+fLmpjFqtJmFhYWTNmjW89fz666/Ez8+PaDQai982btwoWlBevXo1CQsLI2q12rRt2bJlJDo6muj1ekIIISNHjiRvv/02a7/XX3+djB49WtQxmhJK618ms2bN4lSEmIwfP75JC8pCOKN/jXz66aekU6dOVtsj5vllQvuXjbv3p5GHH36YPPnkk1brboq4Y/8aFaGysjIxp9ikccf+NeePP/4gKpWKZGZmWq27KeOO/SyX3OwRrnHmGM1yLVu2BABkZGSgoKAAkydPNpXx9/fH+PHjcfr0aav1hIaGwsfHx672nDlzBuPHj2etqjtlyhTk5eUhMzMTAFBfX4+AgADWfoGBgTh//jw0Go1dx/c0lNa/FHlxZv9WVFSYjsOHmOeXwo8n9Gd8fDxOnz6N8ePHW627KeLO/Ttw4EBERUVh4sSJiImJsVpvOQEqKAAABdlJREFUU8Wd+9fI+vXrMWnSJHTo0MFq3U0Zd+xnueRmj1OECCGYN28exowZg759+wIACgoKAAARERGsshEREabfzCkpKcFHH32El156ye42FRQUcB6b2bYpU6Zg3bp1iIuLAyEEsbGx2LBhAzQaDYqLi+1ug6egxP6lyIcz+zc9PR1ff/015syZY7VNYp5fCjfu3p9t27aFv78/hgwZgrlz5+L555+3WndTw137NyoqCmvXrsX27duxY8cO9OjRAxMnTsTx48et1t3UcNf+ZZKfn4+9e/fSZ9cK7trPcsnNHqcIvfLKK7hy5Qq2bdtm8ZtKpWJ9J4RYbAOAyspKTJs2Db1798bChQslHb9Pnz4IDg5GcHAw7rvvPqvHZm7/4IMPcN9992HEiBHw9fXF9OnTMXv2bACAt7e3pDZ4MkrtX4o8OKt/8/LycO+99+LRRx9lvSBtfX4p3Lh7f544cQKxsbFYs2YNVqxYwXkeTRl37d8ePXrghRdewKBBgzBy5EisXr0a06ZNw+effy7yzJsG7tq/TDZt2oTmzZtzJjGiGHDXfpZLbvYon6BXX30Vu3btwvHjx9G2bVvT9sjISAAGLTIqKsq0vaioyELjrKqqwr333ovg4GD88ccf8PX1ldSGPXv2mExygYGBpuOba9BFRUUAGjXcwMBAbNiwAd999x0KCwtNM1YhISFo3bq1pDZ4KkrtX4o8OKt/8/LyMGHCBIwcORJr165l/Wbr80uxxBP6s1OnTgCAfv36obCwEIsWLcLMmTPFXwQPxhP6l8mIESOwdetWwfNuKnhC/xJCsGHDBjz11FPw8/OTdP5NBXfuZ9nkZkkRRQpFr9eTuXPnkujoaJKWlsb5e2RkJPnkk09M2+rr6y2CvioqKsiIESPI+PHjSU1NjdVjSk2W0Lx5c1JfX2/atnz5csHg3HHjxpGZM2eKOoYno/T+ZUKTJUjHmf2bk5NDunXrRh5//HHRGaKkPr9NvX89rT+NLFmyhHTo0EHUMTwZT+3fGTNmkAkTJog6hifjSf1rTIqRkJAgqu6mhCf1MxNb5GaPUIT+/e9/k7CwMHL06FFWysTa2lpTmeXLl5OwsDCyY8cOkpCQQGbOnMlKA1hZWUmGDx9O+vXrR65fv86bUvPmzZskPj6eLF68mAQHB5P4+HgSHx9PqqqqeNtXXl5OIiIiyMyZM0lCQgLZsWMHCQ0NZaUBTE1NJVu2bCFpaWnk3Llz5LHHHiMtW7YkGRkZ8l8wN0Pp/UsIIUlJSSQ+Pp488MAD5K677jLtx8S4bfDgweSJJ54g8fHxJCkpSb4L5aY4q39zc3NJ165dyd13301ycnJYZawh5vklhPavEU/oz2+++Ybs2rWLpKWlkbS0NLJhwwYSGhpK3nvvPQdcMffCE/r3yy+/JH/88QdJS0sjiYmJZP78+QQA2b59uwOumHvhCf1r5MknnyTDhw+X8ep4Dp7Qz3LJzR6hCAHg/GzcuNFURq/Xk4ULF5LIyEji7+9Pxo0bx5olMM4ccH2YF3XWrFmcZWJiYqy28cqVK2Ts2LHE39+fREZGkkWLFrG02uTkZDJgwAASGBhIQkNDyfTp00lKSopcl8itcYf+7dChA+d+QudBZ5id178bN27kLSOE0PPLdx5NsX89oT9XrlxJ+vTpQ4KCgkhoaCgZOHAgWb16NdHpdLJdJ3fFE/r3k08+IV26dCEBAQGkRYsWZMyYMWT37t2yXSN3xhP6lxCDIB0YGEjWrl0ry3XxNDyhn+WSm1W3LwiFQqFQKBQKhUKhNBk8LmschUKhUCgUCoVCoQhBFSEKhUKhUCgUCoXS5KCKEIVCoVAoFAqFQmlyUEWIQqFQKBQKhUKhNDmoIkShUCgUCoVCoVCaHFQRolAoFAqFQqFQKE0OqghRKBQKhUKhUCiUJgdVhCgUCoVCoVAoFEqTgypCFAqFQqFQKBQKpclBFSEKhUKhUCgUCoXS5KCKEIVCoVAoFAqFQmly/D8BynId96TsRQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1000x500 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax1, ax2, ax3,\n",
    "     ) = plt.subplots(nrows = 3,\n",
    "                      ncols = 1, figsize = [10, 5])\n",
    "ax1.plot(df_train['time'], df_train['y_imp'])\n",
    "ax2.plot(df_train['time'], df_train['y_imp_r2'])\n",
    "ax3.plot(df_train['time'], df_train['y_imp_r3'])\n",
    "\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "34b70287",
   "metadata": {},
   "source": [
    "There is still some significant intra-day variation that results in spikes in the graphs. Let's try to further smooth these spikes by applying log-normalization. Let's then plot the results. Note: when predicting the values, make sure to exponentiate the predictions!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "d9a1b0f9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>temp</th>\n",
       "      <th>dwpt</th>\n",
       "      <th>rhum</th>\n",
       "      <th>prcp</th>\n",
       "      <th>snow</th>\n",
       "      <th>wdir</th>\n",
       "      <th>wspd</th>\n",
       "      <th>wpgt</th>\n",
       "      <th>pres</th>\n",
       "      <th>...</th>\n",
       "      <th>hour</th>\n",
       "      <th>season</th>\n",
       "      <th>h_type</th>\n",
       "      <th>is_holiday</th>\n",
       "      <th>y_imp</th>\n",
       "      <th>y_imp_r2</th>\n",
       "      <th>y_imp_r3</th>\n",
       "      <th>y_imp_log</th>\n",
       "      <th>y_imp_r2_log</th>\n",
       "      <th>y_imp_r3_log</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2021-09-01 00:00:00+03:00</td>\n",
       "      <td>11.2</td>\n",
       "      <td>10.3</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>16.7</td>\n",
       "      <td>1012.6</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>0.577</td>\n",
       "      <td>0.5770</td>\n",
       "      <td>0.577000</td>\n",
       "      <td>-0.549913</td>\n",
       "      <td>-0.549913</td>\n",
       "      <td>-0.549913</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2021-09-01 01:00:00+03:00</td>\n",
       "      <td>10.7</td>\n",
       "      <td>9.6</td>\n",
       "      <td>93.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1012.6</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>0.594</td>\n",
       "      <td>0.5855</td>\n",
       "      <td>0.594000</td>\n",
       "      <td>-0.520876</td>\n",
       "      <td>-0.535289</td>\n",
       "      <td>-0.520876</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2021-09-01 02:00:00+03:00</td>\n",
       "      <td>9.9</td>\n",
       "      <td>9.0</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>320.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1012.2</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>0.685</td>\n",
       "      <td>0.6395</td>\n",
       "      <td>0.618667</td>\n",
       "      <td>-0.378336</td>\n",
       "      <td>-0.447069</td>\n",
       "      <td>-0.480189</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2021-09-01 03:00:00+03:00</td>\n",
       "      <td>10.0</td>\n",
       "      <td>8.4</td>\n",
       "      <td>90.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>330.0</td>\n",
       "      <td>7.2</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1011.9</td>\n",
       "      <td>...</td>\n",
       "      <td>3</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>1.016</td>\n",
       "      <td>0.8505</td>\n",
       "      <td>0.765000</td>\n",
       "      <td>0.015873</td>\n",
       "      <td>-0.161931</td>\n",
       "      <td>-0.267879</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2021-09-01 04:00:00+03:00</td>\n",
       "      <td>9.0</td>\n",
       "      <td>8.1</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>300.0</td>\n",
       "      <td>3.6</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1011.4</td>\n",
       "      <td>...</td>\n",
       "      <td>4</td>\n",
       "      <td>Autumn</td>\n",
       "      <td>not_hol</td>\n",
       "      <td>0</td>\n",
       "      <td>0.677</td>\n",
       "      <td>0.8465</td>\n",
       "      <td>0.792667</td>\n",
       "      <td>-0.390084</td>\n",
       "      <td>-0.166645</td>\n",
       "      <td>-0.232352</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 27 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                       time  temp  dwpt  rhum  prcp  snow   wdir  wspd  wpgt  \\\n",
       "0 2021-09-01 00:00:00+03:00  11.2  10.3  94.0   0.0   0.0  320.0   7.2  16.7   \n",
       "1 2021-09-01 01:00:00+03:00  10.7   9.6  93.0   0.0   0.0  320.0   7.2  13.0   \n",
       "2 2021-09-01 02:00:00+03:00   9.9   9.0  94.0   0.0   0.0  320.0   7.2  13.0   \n",
       "3 2021-09-01 03:00:00+03:00  10.0   8.4  90.0   0.0   0.0  330.0   7.2  13.0   \n",
       "4 2021-09-01 04:00:00+03:00   9.0   8.1  94.0   0.0   0.0  300.0   3.6  13.0   \n",
       "\n",
       "     pres  ... hour  season   h_type  is_holiday  y_imp y_imp_r2  y_imp_r3  \\\n",
       "0  1012.6  ...    0  Autumn  not_hol           0  0.577   0.5770  0.577000   \n",
       "1  1012.6  ...    1  Autumn  not_hol           0  0.594   0.5855  0.594000   \n",
       "2  1012.2  ...    2  Autumn  not_hol           0  0.685   0.6395  0.618667   \n",
       "3  1011.9  ...    3  Autumn  not_hol           0  1.016   0.8505  0.765000   \n",
       "4  1011.4  ...    4  Autumn  not_hol           0  0.677   0.8465  0.792667   \n",
       "\n",
       "   y_imp_log y_imp_r2_log y_imp_r3_log  \n",
       "0  -0.549913    -0.549913    -0.549913  \n",
       "1  -0.520876    -0.535289    -0.520876  \n",
       "2  -0.378336    -0.447069    -0.480189  \n",
       "3   0.015873    -0.161931    -0.267879  \n",
       "4  -0.390084    -0.166645    -0.232352  \n",
       "\n",
       "[5 rows x 27 columns]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_train['y_imp_log'] = np.log(df_train['y_imp'])\n",
    "df_train['y_imp_r2_log'] =  np.log(df_train['y_imp_r2'])\n",
    "df_train['y_imp_r3_log'] = np.log(df_train['y_imp_r3'])\n",
    "df_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "5e9b52f0",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x500 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax1, ax2, ax3) = plt.subplots(nrows = 3, \n",
    "                      ncols = 1, figsize = [10, 5])\n",
    "ax1.plot(df_train['time'], df_train['y_imp_log'])\n",
    "ax2.plot(df_train['time'], df_train['y_imp_r2_log'])\n",
    "ax3.plot(df_train['time'], df_train['y_imp_r3_log'])\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0f7faa08",
   "metadata": {},
   "source": [
    "Right, now we see that graphs are smoother - with the exception of a spike sometime in May. This is likely because on a particular timepoint, the consumption was zero. Could be a power outage. We'll keep this presently as is but we acknowledge that it may be reasonable to apply additional transformations (e.g., average this, etc)."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "92ebda8c",
   "metadata": {},
   "source": [
    "## Preparing the Training, Validation, and Test Data\n",
    "Here, we split the initial training data to training and validation sets. We also keep the test data separately."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "7dcbe792",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Extract all y-variables (including the augmented/transformed ones) - NB! There are several y-variables!\n",
    "y_features = df_train.filter(regex='y_')\n",
    "\n",
    "# Prepare the X-data\n",
    "X = df_train.drop(columns = df_train.filter(regex='y_'))\n",
    "X =  X.drop(columns = ['time'])\n",
    "\n",
    "# Find the split time index (1st May 2022)\n",
    "split_idx = X[(X['month_name'] == 'May') & (X['year'] == 2022) &  (X['hour'] == 0)].index[0]\n",
    "X = X[split_idx:].reset_index(drop = True)\n",
    "y_features = y_features[split_idx:].reset_index(drop = True)\n",
    "y = y_features['y_imp_r3'] # Extract the three-hour moving average imputed y\n",
    "\n",
    "# Test data\n",
    "X_test = df_test.drop(columns = ['time'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "fda729d5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2228, 556)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train-validation split\n",
    "split_index = math.ceil(len(X) * 0.8) # 80%-20% split\n",
    "\n",
    "train_X = X[:split_index].reset_index(drop = True)\n",
    "val_X = X[split_index:].reset_index(drop = True)\n",
    "train_y = y[:split_index].reset_index(drop = True)\n",
    "val_y = y[split_index:].reset_index(drop = True)\n",
    "\n",
    "len(train_X), len(val_X)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f9555339",
   "metadata": {},
   "source": [
    "### Feature pre-processing\n",
    "The cell below includes the feature preprocessing. We use the `MinMaxScaler()` to scale the numerical features, while One-Hot-Encoding OHE) is used for caterical features. The former is first fitted on the training split and then, subsequently, the transformations are made based on that fit. The OHE is done with training, validation, and test set combined, since this would ensure the existence of all categories in all datasets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "96d50485",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training data dims, X: ((2228, 43),), y = (2228,).\n",
      "Validation data dims, X: ((556, 43),), y = (556,).\n",
      "Test data dims, X: ((556, 43),).\n"
     ]
    }
   ],
   "source": [
    "# Collect lists of variable names based on type for subsequent use\n",
    "num_vars = []\n",
    "cat_vars = []\n",
    "int_vars = []\n",
    "for i in range(len(train_X.dtypes)):\n",
    "    if train_X.dtypes.values[i] == 'float64':\n",
    "        num_vars.append(train_X.dtypes.index[i])\n",
    "    elif train_X.dtypes.values[i] == 'object':\n",
    "        cat_vars.append(train_X.dtypes.index[i])\n",
    "    elif train_X.dtypes.values[i] == 'int64':\n",
    "        int_vars.append(train_X.dtypes.index[i])\n",
    "\n",
    "# Scale the float variables\n",
    "scaler = MinMaxScaler()\n",
    "\n",
    "# Fit on the training set\n",
    "scaler.fit(train_X[num_vars])\n",
    "# Transform the features of other data sets (fit based on training ste)\n",
    "train_X[num_vars] = scaler.transform(train_X[num_vars])\n",
    "val_X[num_vars] = scaler.transform(val_X[num_vars])\n",
    "X_test[num_vars] = scaler.transform(X_test[num_vars])\n",
    "\n",
    "# One-hot encoding for categorical vars\n",
    "## Should be done for train + val + test in one df because otherwise N of vars not the same\n",
    "ohe_vars = pd.get_dummies(pd.concat([train_X[cat_vars], \n",
    "                                     val_X[cat_vars],\n",
    "                                     X_test[cat_vars]\n",
    "                                    ], \n",
    "                                    ignore_index=True))\n",
    "ohe_vars = pd.get_dummies(pd.concat([train_X[cat_vars], val_X[cat_vars], X_test[cat_vars]], ignore_index=True))\n",
    "\n",
    "# Drop the old categorical features\n",
    "train_X = train_X.drop(columns = cat_vars)\n",
    "val_X = val_X.drop(columns = cat_vars)\n",
    "X_test = X_test.drop(columns = cat_vars)\n",
    "\n",
    "# Add the OHE variables to train, test, and val data\n",
    "## lengths of datasets for index-based extraction\n",
    "len_xtrain = len(train_X)\n",
    "len_xval = len(val_X)\n",
    "len_xtest = len(X_test)\n",
    "\n",
    "# Add the OHE vars to training, validation, and test sets instead of the previous categorical vars\n",
    "train_X = train_X.join(ohe_vars[:len_xtrain])\n",
    "val_X = val_X.join(ohe_vars[len_xtrain:(len_xtrain + len_xval)].reset_index(drop = True))\n",
    "X_test = X_test.join(ohe_vars[(len_xtrain + len_xval):].reset_index(drop = True))\n",
    "\n",
    "print(f'Training data dims, X: {train_X.shape,}, y = {train_y.shape}.')\n",
    "print(f'Validation data dims, X: {val_X.shape,}, y = {val_y.shape}.')\n",
    "print(f'Test data dims, X: {val_X.shape,}.')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "68cfe0a8",
   "metadata": {},
   "source": [
    "# Modeling\n",
    "In this section, we are running the models we used for predictions throughout the duration of this project. Of note, most of the modeling was performed in other notebooks - we refer the reader to look into them for more details. \n",
    "\n",
    "Because we adapted our approaches, data preprocessing, and modeling throughout the project time, the notebooks - although using largely similar approaches - have variations that result in differences between solutions below. For instance, there may have been variations in data cleaning, target variable use (i.e., using original,or imputed, or a moving average-based, or log-normalized values), training-validation data splits, etc. We duly acknowledge that the present notebooks serves as a general summary of our workflow, and that some differences in model performance rely on differences in methods from the present notebook. However, we also want to stress that when working with these mdoels, we discovered that the general *relative* goodness of models against each other remains roughly in the same order.\n",
    "\n",
    "Below, we are presenting variations of the better solutions tom illustrate our workflow outlined in Figure 1.\n",
    "\n",
    "As mentioned above, we found, in general, that the best models are obtained when we use the three-day moving average (`y_imp_r3`) as the target variable. During trying out different models and timeframes, we discovered that the models with data from May 1st 2022 onward performed well (as opposed to using all available data from the apst). Finally, though alternative solutions are present, the meaningful way to train the models would be by using trian-validation data splits based on time, i.e., validation data should be later in time than training data.\n",
    "\n",
    "In general, we used three approached: (a) algorithms that do not model time-series (predicting the traget from features), (b) algorithms/frameworks designed for time-series forecasts (forecasting the target from a sequence of itself), and (c) ensemble predictions of the best of two approaches (averaging the predictions of the best models' predictions). \n",
    "\n",
    "Before we start modeling, we create a data table that would store the name, runtime, and MAE of each model. Additionally, a dataframe that includes an array for every model's predictions is created (in LR code cell) and updated. This would allow later for a simple ensembling."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "6c5be942",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>runtime</th>\n",
       "      <th>MAE</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: [model, runtime, MAE]\n",
       "Index: []"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Initialize a df where the model performance will be collected\n",
    "df_models = pd.DataFrame(columns = ['model', 'runtime', 'MAE'])\n",
    "df_models.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "28226956",
   "metadata": {},
   "source": [
    "## Non-Time-Series Based Approaches\n",
    "In the present section, we tried using the mdoeling approach which is not taking into account the nature of the data (which is time-series). To that end, we used the training-validation data splits (see above) made on the Kaggle training data where last observations were predicted from the first ones. \n",
    "\n",
    "Here, we have regression-based algorithms (Linear Regression, Lasso, Ridge, Elastic Net), as well as tree-based (Decision Tree) and ensemble (Random Forest, XGBoost) algorithms."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b5e698e1",
   "metadata": {},
   "source": [
    "### Regressions\n",
    "#### Linear Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "49844639",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.31954576312699356\n"
     ]
    }
   ],
   "source": [
    "lambda_ = 1 # provide a lambda value\n",
    "\n",
    "start = time.time()\n",
    "lr = LinearRegression()\n",
    "lr.fit(train_X, train_y)\n",
    "runtime = time.time() - start\n",
    "\n",
    "# Predictions\n",
    "predict_LR = lr.predict(val_X)\n",
    "mae = mean_absolute_error(val_y, predict_LR)\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "# Append to models df\n",
    "df_models = df_models.append({'model': 'LR', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Another df for model predictions\n",
    "df_preds = pd.DataFrame({'LR': predict_LR})"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "48b895c0",
   "metadata": {},
   "source": [
    "#### Lasso Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "f63b84dc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.277727790628636\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "lr_lasso = Lasso(lambda_)\n",
    "lr_lasso.fit(train_X, train_y)\n",
    "runtime = time.time() - start\n",
    "\n",
    "# Predictions\n",
    "predict_LASSO = lr_lasso.predict(val_X)\n",
    "mae = mean_absolute_error(val_y, predict_LASSO)\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "# Append to models df\n",
    "df_models = df_models.append({'model': 'LASSO', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Add the predictions\n",
    "df_preds['LASSO'] = predict_LASSO"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "36bafac8",
   "metadata": {},
   "source": [
    "#### Ridge Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "73544199",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.28843628314592157\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "lr_ridge = Ridge(lambda_)\n",
    "lr_ridge.fit(train_X, train_y)\n",
    "runtime = time.time() - start\n",
    "\n",
    "# Predictions\n",
    "predict_RIDGE = lr_ridge.predict(val_X)\n",
    "mae = mean_absolute_error(val_y, predict_RIDGE)\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "# Append to models df\n",
    "df_models = df_models.append({'model': 'RIDGE', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Add the predictions\n",
    "df_preds['RIDGE'] = predict_RIDGE"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0cf60c9e",
   "metadata": {},
   "source": [
    "#### Elastic Net Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "1d8f4f8d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.2509215909038105\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "lr_elastic = ElasticNet(alpha = lambda_)\n",
    "lr_elastic.fit(train_X, train_y)\n",
    "runtime = time.time() - start\n",
    "\n",
    "# Predictions\n",
    "predict_ElasticNet = lr_elastic.predict(val_X)\n",
    "mae = mean_absolute_error(val_y, predict_ElasticNet)\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "df_models = df_models.append({'model': 'ElasticNet', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Add the predictions\n",
    "df_preds['ElasticNet'] = predict_ElasticNet"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7d8ff09b",
   "metadata": {},
   "source": [
    "### Intermediate Summary: Regressions\n",
    "Below we are displaying the model MAEs with model runtimes for the regression models:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "ed88090a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>runtime</th>\n",
       "      <th>MAE</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ElasticNet</td>\n",
       "      <td>0.003238</td>\n",
       "      <td>0.250922</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>LASSO</td>\n",
       "      <td>0.004800</td>\n",
       "      <td>0.277728</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>RIDGE</td>\n",
       "      <td>0.003665</td>\n",
       "      <td>0.288436</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>LR</td>\n",
       "      <td>0.007286</td>\n",
       "      <td>0.319546</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        model   runtime       MAE\n",
       "0  ElasticNet  0.003238  0.250922\n",
       "1       LASSO  0.004800  0.277728\n",
       "2       RIDGE  0.003665  0.288436\n",
       "3          LR  0.007286  0.319546"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_models.sort_values('MAE', ascending = True).reset_index(drop = True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "00b4fd2d",
   "metadata": {},
   "source": [
    "Here, Elastic Net Regression seems to perform the best."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2e668e26",
   "metadata": {},
   "source": [
    "### Decision Tree\n",
    "We also tried the decision tree regressor. After fitting a baseline, i.e., not tuned model, we used grid search for tuning the hyperparameters. This resulted in the below-outlined model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "a4f39761",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.24506364901985161\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "tuned_dt = DecisionTreeRegressor(max_depth=6,\n",
    "                                 max_features='auto',\n",
    "                                 max_leaf_nodes=20,\n",
    "                                 min_samples_leaf=3,\n",
    "                                 min_weight_fraction_leaf=0.1,\n",
    "                                 splitter='random', \n",
    "                                random_state=0)\n",
    "tuned_dt.fit(train_X, train_y)\n",
    "runtime = time.time() - start\n",
    "\n",
    "# Predictions\n",
    "predict_DT = tuned_dt.predict(val_X)\n",
    "mae = mean_absolute_error(val_y, predict_DT)\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "df_models = df_models.append({'model': 'Tuned DT', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Add the predictions\n",
    "df_preds['Tuned DT'] = predict_DT"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dd982239",
   "metadata": {},
   "source": [
    "### Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "1699a5df",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.22201337687744543\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "tuned_rf = RandomForestRegressor(n_estimators = 950, \n",
    "                                 bootstrap = False, \n",
    "                                 max_depth = None, \n",
    "                                 max_features = 'sqrt', \n",
    "                                 min_samples_leaf = 1, \n",
    "                                 min_samples_split = 2,\n",
    "                                random_state= 0)\n",
    "\n",
    "tuned_rf.fit(train_X, train_y)\n",
    "runtime = time.time() - start\n",
    "\n",
    "# Predictions\n",
    "predict_RF = tuned_rf.predict(val_X)\n",
    "mae = mean_absolute_error(val_y, predict_RF)\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "df_models = df_models.append({'model': 'Tuned RF', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Add the predictions\n",
    "df_preds['Tuned RF'] = predict_RF"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4b7200ab",
   "metadata": {},
   "source": [
    "### XGBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "33696e89",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.21909084500967058\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "xgb_best = xgb.XGBRegressor(colsample_bytree = 0.9, \n",
    "                                 learning_rate = 0.01, \n",
    "                                 max_depth = 10, \n",
    "                                 n_estimators = 400, seed=0)\n",
    "xgb_best.fit(train_X, train_y)\n",
    "runtime = time.time() - start\n",
    "\n",
    "# Predictions\n",
    "predict_XGBoost = xgb_best.predict(val_X)\n",
    "mae = mean_absolute_error(val_y, predict_XGBoost)\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "df_models = df_models.append({'model': 'XGBoost', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Add the predictions\n",
    "df_preds['XGBoost'] = predict_XGBoost"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "410733be",
   "metadata": {},
   "source": [
    "### Intermediate summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "bfd43695",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>runtime</th>\n",
       "      <th>MAE</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>XGBoost</td>\n",
       "      <td>1.595126</td>\n",
       "      <td>0.219091</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Tuned RF</td>\n",
       "      <td>3.466644</td>\n",
       "      <td>0.222013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Tuned DT</td>\n",
       "      <td>0.003483</td>\n",
       "      <td>0.245064</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ElasticNet</td>\n",
       "      <td>0.003238</td>\n",
       "      <td>0.250922</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>LASSO</td>\n",
       "      <td>0.004800</td>\n",
       "      <td>0.277728</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>RIDGE</td>\n",
       "      <td>0.003665</td>\n",
       "      <td>0.288436</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>LR</td>\n",
       "      <td>0.007286</td>\n",
       "      <td>0.319546</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        model   runtime       MAE\n",
       "0     XGBoost  1.595126  0.219091\n",
       "1    Tuned RF  3.466644  0.222013\n",
       "2    Tuned DT  0.003483  0.245064\n",
       "3  ElasticNet  0.003238  0.250922\n",
       "4       LASSO  0.004800  0.277728\n",
       "5       RIDGE  0.003665  0.288436\n",
       "6          LR  0.007286  0.319546"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_models.sort_values('MAE').reset_index(drop = True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b13c6f59",
   "metadata": {},
   "source": [
    "## Time-Series Based Approaches\n",
    "In this section, we switch to time-series based forecasting. To that end, we try out two frameworks: the `prophet` library and `skforecast`. Both take the taraget variable as the input (therefore, other fetaures are ignored completely). Additionally, `prophet` takes the timestamps as input. As mentioned above, we are using the 3-hour moving averages as the target variable."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fc8e6b8c",
   "metadata": {},
   "source": [
    "### The `Prophet` Approach\n",
    "First, data needs to be converted into the specific format. Here we use the entire Kaggle training set. After fitting the model, we predict the test data values."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "16b38f92",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_ts = df_train[['time', 'y_imp_r3']] # timeseries df\n",
    "df_prophet = df_ts[split_idx:].reset_index(drop = True) # use the split index starting from May 1, 2022\n",
    "df_prophet = df_prophet.rename(columns = {'time':'ds', 'y_imp_r3':'y'})\n",
    "df_prophet['ds'] = df_prophet['ds'].dt.tz_localize(None)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "de2e008a",
   "metadata": {},
   "source": [
    "Then, the forecasting can be applied."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "ee69eb7a",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "prophet    INFO  Disabling yearly seasonality. Run prophet with yearly_seasonality=True to override this.\n",
      "cmdstanpy  DEBUG input tempfile: /var/folders/28/mdvjx70j6qqfykv1xw8gz1n00000gn/T/tmpptkan_38/ps09j2j2.json\n",
      "cmdstanpy  DEBUG input tempfile: /var/folders/28/mdvjx70j6qqfykv1xw8gz1n00000gn/T/tmpptkan_38/cdnwpg7m.json\n",
      "cmdstanpy  DEBUG idx 0\n",
      "cmdstanpy  DEBUG running CmdStan, num_threads: None\n",
      "cmdstanpy  DEBUG CmdStan args: ['/Users/drdr/opt/anaconda3/envs/ml_2022/lib/python3.7/site-packages/prophet/stan_model/prophet_model.bin', 'random', 'seed=5919', 'data', 'file=/var/folders/28/mdvjx70j6qqfykv1xw8gz1n00000gn/T/tmpptkan_38/ps09j2j2.json', 'init=/var/folders/28/mdvjx70j6qqfykv1xw8gz1n00000gn/T/tmpptkan_38/cdnwpg7m.json', 'output', 'file=/var/folders/28/mdvjx70j6qqfykv1xw8gz1n00000gn/T/tmpptkan_38/prophet_model63tosftj/prophet_model-20221211201413.csv', 'method=optimize', 'algorithm=lbfgs', 'iter=10000']\n",
      "20:14:13 - cmdstanpy - INFO - Chain [1] start processing\n",
      "cmdstanpy  INFO  Chain [1] start processing\n",
      "20:14:13 - cmdstanpy - INFO - Chain [1] done processing\n",
      "cmdstanpy  INFO  Chain [1] done processing\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "m = Prophet(changepoint_prior_scale=0.01).fit(df_prophet)\n",
    "future = m.make_future_dataframe(periods=168, freq='H')\n",
    "fcst = m.predict(future)\n",
    "runtime = time.time() - start\n",
    "\n",
    "# fig = m.plot(fcst) # Possibility to display the forecast figure\n",
    "# fcst.head() # uncomment to see how the forecasting output df looks like"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e502b229",
   "metadata": {},
   "source": [
    "Let's find the predictions for the timeframe that was also used for other models and compute the MAE."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "78f187b3",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.2166286217409674\n"
     ]
    }
   ],
   "source": [
    "predict_Prophet = fcst.iloc[len(train_X):len(train_X)+len(val_X)].reset_index(drop = True)['yhat']\n",
    "\n",
    "mae = mean_absolute_error(val_y, predict_Prophet)\n",
    "\n",
    "print(f'MAE: {mae}')\n",
    "df_models = df_models.append({'model': 'Prophet', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Add the predictions\n",
    "df_preds['Prophet'] = predict_Prophet\n",
    "\n",
    "# Extract the predictions for test data (Kaggle)\n",
    "# predictions = fcst.iloc[-168:][['ds', 'yhat']].reset_index(drop = True)\n",
    "# predictions.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6e569249",
   "metadata": {},
   "source": [
    "## `skforecast` Approaches\n",
    "### Random Forest Regressor \n",
    "\n",
    "Here, we need to define how mayn lags we should use for predictions. In our case, we already saw that a three-hour lag could provide the best results. Hence, below we use the original imputed target variable with three logs. Additionally, here we use the entire Kaggle training set. After fitting the model, we predict the test data values."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "26ddac4c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.30249139917594053\n"
     ]
    }
   ],
   "source": [
    "# See this tutorial: https://www.cienciadedatos.net/documentos/py27-time-series-forecasting-python-scikitlearn.html\n",
    "\n",
    "start = time.time()\n",
    "skforecast_RF = ForecasterAutoreg(\n",
    "                regressor = RandomForestRegressor(n_estimators = 500, random_state=0),\n",
    "                lags = 1\n",
    "             )\n",
    "skforecast_RF.fit(y = train_y)\n",
    "runtime = time.time() - start\n",
    "\n",
    "steps = len(val_y) # Number of predicitons/test data size\n",
    "predict_skforecast_RF = skforecast_RF.predict(steps=steps)\n",
    "\n",
    "mae = mean_absolute_error(val_y, predict_skforecast_RF)\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "df_models = df_models.append({'model': 'skforecast RF', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Add the predictions\n",
    "df_preds['skforecast RF'] = predict_skforecast_RF"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5d8a4385",
   "metadata": {},
   "source": [
    "### XGBoost Regressor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "d0509315",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.2742885342241689\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "skforecast_XGBoost = ForecasterAutoreg(\n",
    "                regressor = xgb.XGBRegressor(colsample_bytree = 0.9, \n",
    "                                 learning_rate = 0.01, \n",
    "                                 max_depth = 10, \n",
    "                                 n_estimators = 400, seed=0),\n",
    "                lags = 1\n",
    "             )\n",
    "\n",
    "skforecast_XGBoost.fit(y = train_y)\n",
    "runtime = time.time() - start\n",
    "\n",
    "steps = len(val_y) # Number of predicitons/test data size\n",
    "predict_skforecast_XGBoost = skforecast_XGBoost.predict(steps=steps)\n",
    "\n",
    "mae = mean_absolute_error(val_y, predict_skforecast_XGBoost)\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "df_models = df_models.append({'model': 'skforecast XGBoost', 'MAE': mae, 'runtime':runtime}, ignore_index = True)\n",
    "\n",
    "# Add the predictions\n",
    "df_preds['skforecast XGBoost'] = predict_skforecast_XGBoost"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "08b5d65b",
   "metadata": {},
   "source": [
    "## Ensembling the TOP Models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "4a078897",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>runtime</th>\n",
       "      <th>MAE</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Prophet</td>\n",
       "      <td>0.733893</td>\n",
       "      <td>0.216629</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>XGBoost</td>\n",
       "      <td>1.595126</td>\n",
       "      <td>0.219091</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Tuned RF</td>\n",
       "      <td>3.466644</td>\n",
       "      <td>0.222013</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      model   runtime       MAE\n",
       "0   Prophet  0.733893  0.216629\n",
       "1   XGBoost  1.595126  0.219091\n",
       "2  Tuned RF  3.466644  0.222013"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Select how may models form the 'Top Models'\n",
    "top_n = 3 \n",
    "\n",
    "# Display the top N models\n",
    "top_models = df_models.sort_values('MAE')[:top_n].reset_index(drop = True)\n",
    "top_models.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "f160f80a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 0.20403362277466705\n"
     ]
    }
   ],
   "source": [
    "# Compute the average of top model predictions\n",
    "df_preds['TOP_ensemble'] = np.mean(df_preds[top_models['model']], axis = 1)\n",
    "\n",
    "# Test the average against the validation data\n",
    "mae = mean_absolute_error(val_y, df_preds['TOP_ensemble'])\n",
    "print(f'MAE: {mae}')\n",
    "\n",
    "# Sum the runtime of individual regressors for the ensemble\n",
    "runtime = top_models['runtime'].sum()\n",
    "\n",
    "df_models = df_models.append({'model': 'TOP Ensemble', 'MAE': mae, 'runtime':runtime}, ignore_index = True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "50186090",
   "metadata": {},
   "source": [
    "## Final Performance Results on Training Data\n",
    "Below are the results of model performances in a tabular format, and as a scatter plot where we show the potential association between model runtime and goodness of predictions. Although the data set for the rpesent project was for one household, modelling thousands or hundreds of thousands of households may increase the demand in resources needed for computation. Hence, the plot could be useful for decision-making when one needs to consider the runtime-accurayc tradeoff."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "a529e13f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>runtime</th>\n",
       "      <th>MAE</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>TOP Ensemble</td>\n",
       "      <td>5.795662</td>\n",
       "      <td>0.204034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Prophet</td>\n",
       "      <td>0.733893</td>\n",
       "      <td>0.216629</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>XGBoost</td>\n",
       "      <td>1.595126</td>\n",
       "      <td>0.219091</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Tuned RF</td>\n",
       "      <td>3.466644</td>\n",
       "      <td>0.222013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Tuned DT</td>\n",
       "      <td>0.003483</td>\n",
       "      <td>0.245064</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>ElasticNet</td>\n",
       "      <td>0.003238</td>\n",
       "      <td>0.250922</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>skforecast XGBoost</td>\n",
       "      <td>0.542847</td>\n",
       "      <td>0.274289</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>LASSO</td>\n",
       "      <td>0.004800</td>\n",
       "      <td>0.277728</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>RIDGE</td>\n",
       "      <td>0.003665</td>\n",
       "      <td>0.288436</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>skforecast RF</td>\n",
       "      <td>1.098848</td>\n",
       "      <td>0.302491</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>LR</td>\n",
       "      <td>0.007286</td>\n",
       "      <td>0.319546</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 model   runtime       MAE\n",
       "0         TOP Ensemble  5.795662  0.204034\n",
       "1              Prophet  0.733893  0.216629\n",
       "2              XGBoost  1.595126  0.219091\n",
       "3             Tuned RF  3.466644  0.222013\n",
       "4             Tuned DT  0.003483  0.245064\n",
       "5           ElasticNet  0.003238  0.250922\n",
       "6   skforecast XGBoost  0.542847  0.274289\n",
       "7                LASSO  0.004800  0.277728\n",
       "8                RIDGE  0.003665  0.288436\n",
       "9        skforecast RF  1.098848  0.302491\n",
       "10                  LR  0.007286  0.319546"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_models = df_models.sort_values('MAE').reset_index(drop = True)\n",
    "df_models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "c94ba5f8",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize = [10, 4])\n",
    "\n",
    "for i in range(len(df_models)):\n",
    "\n",
    "    plt.scatter(df_models['runtime'][i], \n",
    "            df_models['MAE'][i], \n",
    "            label = df_models['model'][i],\n",
    "               marker = 'x',\n",
    "               s = 60)\n",
    "    plt.legend()\n",
    "plt.xlabel('Runtime (sec)')\n",
    "plt.ylabel('MAE')\n",
    "plt.title('MAE and Runtime for Each Regressor')\n",
    "plt.show();"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}