224_Predicting_equity_premium.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# A Machine Learning Approach to Equity Premium Prediction\n",
    "By Shane Johnson (ND CS), Catherine Markley (ND CS), Yuanhao Niu (ND Econ), Nick Rocco (ND CS) \n",
    "<br>-- Fall 2018 CSE 40647/60647 Data Science Project"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Introduction\n",
    "We try to use machine learning methods to predict equity premium based on predictors from the seminal paper of Goyal and Welch (2008). \n",
    "\n",
    "We first build and clean our dataset. Then we try to predict bull/bear market as a classification job. Portfolio based on the prediction were constructed accordingly. In a parallel task, we use different regression approaches to predict equity premium. Our models outperforme the historical average forecast. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Cleaning\n",
    "\n",
    "Raw data are downloaded from Prof. Goyal's webside. We included their orginal monthly data from 1926/12 to 2013/12."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "df = pd.read_csv(\"raw_predictors_until2013.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Generate the predictors based on the raw data, as in Welch and Goyal (2008)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# equity premium, i.e., the total rate of return on the stock market minus the prevailing short-term interest rate.\n",
    "df['premium'] = df['CRSP_SPvw'] - df['Rfree']\n",
    "\n",
    "# Dividend Price Ratio (d/p) is the difference between the log of dividends and the log of prices. \n",
    "df['d/p'] = np.log(df['D12']) - np.log(df['Index'])\n",
    "\n",
    "# Dividend Yield (d/y) is the difference between the log of dividends and the log of lagged prices.\n",
    "df['d/y'] = np.log(df['D12'])- np.log(df['Index'].shift(1))\n",
    "\n",
    "# Earnings Price Ratio (e/p) is the difference between the log of earnings and the log of prices.\n",
    "df['e/p'] = np.log(df['E12']) - np.log(df['Index'])\n",
    "\n",
    "# Dividend Payout Ratio (d/e) is the difference between the log of dividends and the log of earnings.\n",
    "df['d/e'] = np.log(df['D12']) - np.log(df['E12'])\n",
    "\n",
    "# Term Spread (tms) is the difference between the long term yield on government bonds and the T-bill.\n",
    "df['tms'] = df['lty'] - df['tbl']\n",
    "\n",
    "# Default Yield Spread (dfy): is the difference between BAA- and AAA- rated cor- porate bond yields.\n",
    "df['dfy'] = df['BAA'] - df['AAA']\n",
    "\n",
    "# Default Return Spread (dfr): is the difference between the return on long-term corporate bonds and returns on the long-term government bonds.\n",
    "df['dfr'] = df['corpr'] - df['ltr']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# drop the first observation for 1926/12 (used it to calculate lagged value)\n",
    "df = df[1:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We generate indicators for bull market."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "sLength = len(df['yyyymm'])\n",
    "# add the bullMarket column with random ints (to be replaced in next cell)\n",
    "df['bull'] = pd.Series(np.random.randn(sLength), index=df.index)\n",
    "# Create bullMarkert for ret\n",
    "bull = []\n",
    "for d in df['CRSP_SPvw']:\n",
    "    if d > 0:\n",
    "        bull.append(1)\n",
    "    else:\n",
    "        bull.append(0)\n",
    "df['bull'] = bull"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop the redundent variables\n",
    "df = df.drop(columns=['Index','D12', 'E12', 'AAA', 'BAA', 'CRSP_SPvwx', 'corpr','Rfree', 'CRSP_SPvw'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The Cross-Sectional Premium (csp) was only available from May 1937 to December 2002. \n",
    "# We compute the average of the data and fill the missing values with the average value.\n",
    "pd.options.mode.chained_assignment=None\n",
    "df['csp'][0:124] = df['csp'][124:911].mean()\n",
    "df['csp'][912:]  = df['csp'][124:911].mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We calculate the benchmark (historical average) prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "premium = 0\n",
    "count = 0\n",
    "\n",
    "temp = df['premium']\n",
    "\n",
    "his_avg = []\n",
    "\n",
    "for i in temp:\n",
    "    premium += i\n",
    "    count += 1\n",
    "    his_avg.append(premium/count)\n",
    "\n",
    "df['his_avg'] = his_avg"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we had the clean dataset to work on!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>yyyymm</th>\n",
       "      <th>Month</th>\n",
       "      <th>his_avg</th>\n",
       "      <th>bull</th>\n",
       "      <th>premium</th>\n",
       "      <th>d/p</th>\n",
       "      <th>d/y</th>\n",
       "      <th>e/p</th>\n",
       "      <th>d/e</th>\n",
       "      <th>svar</th>\n",
       "      <th>csp</th>\n",
       "      <th>b/m</th>\n",
       "      <th>ntis</th>\n",
       "      <th>tbl</th>\n",
       "      <th>lty</th>\n",
       "      <th>ltr</th>\n",
       "      <th>tms</th>\n",
       "      <th>dfy</th>\n",
       "      <th>dfr</th>\n",
       "      <th>infl</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>192701</td>\n",
       "      <td>1</td>\n",
       "      <td>-0.005602</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.005602</td>\n",
       "      <td>-2.942374</td>\n",
       "      <td>-2.963349</td>\n",
       "      <td>-2.374773</td>\n",
       "      <td>-0.567601</td>\n",
       "      <td>0.000470</td>\n",
       "      <td>0.00037</td>\n",
       "      <td>0.443706</td>\n",
       "      <td>0.050834</td>\n",
       "      <td>0.0323</td>\n",
       "      <td>0.0351</td>\n",
       "      <td>0.0075</td>\n",
       "      <td>0.0028</td>\n",
       "      <td>0.0095</td>\n",
       "      <td>-0.0019</td>\n",
       "      <td>-0.011299</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>192702</td>\n",
       "      <td>2</td>\n",
       "      <td>0.018589</td>\n",
       "      <td>1</td>\n",
       "      <td>0.042780</td>\n",
       "      <td>-2.979535</td>\n",
       "      <td>-2.932946</td>\n",
       "      <td>-2.430353</td>\n",
       "      <td>-0.549182</td>\n",
       "      <td>0.000287</td>\n",
       "      <td>0.00037</td>\n",
       "      <td>0.428501</td>\n",
       "      <td>0.051682</td>\n",
       "      <td>0.0329</td>\n",
       "      <td>0.0347</td>\n",
       "      <td>0.0088</td>\n",
       "      <td>0.0018</td>\n",
       "      <td>0.0092</td>\n",
       "      <td>-0.0019</td>\n",
       "      <td>-0.005714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>192703</td>\n",
       "      <td>3</td>\n",
       "      <td>0.013945</td>\n",
       "      <td>1</td>\n",
       "      <td>0.004657</td>\n",
       "      <td>-2.976535</td>\n",
       "      <td>-2.970053</td>\n",
       "      <td>-2.445079</td>\n",
       "      <td>-0.531456</td>\n",
       "      <td>0.000924</td>\n",
       "      <td>0.00037</td>\n",
       "      <td>0.469765</td>\n",
       "      <td>0.046370</td>\n",
       "      <td>0.0320</td>\n",
       "      <td>0.0331</td>\n",
       "      <td>0.0253</td>\n",
       "      <td>0.0011</td>\n",
       "      <td>0.0092</td>\n",
       "      <td>-0.0170</td>\n",
       "      <td>-0.005747</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>192704</td>\n",
       "      <td>4</td>\n",
       "      <td>0.013008</td>\n",
       "      <td>1</td>\n",
       "      <td>0.010196</td>\n",
       "      <td>-2.984225</td>\n",
       "      <td>-2.967143</td>\n",
       "      <td>-2.471309</td>\n",
       "      <td>-0.512916</td>\n",
       "      <td>0.000603</td>\n",
       "      <td>0.00037</td>\n",
       "      <td>0.456754</td>\n",
       "      <td>0.050518</td>\n",
       "      <td>0.0339</td>\n",
       "      <td>0.0333</td>\n",
       "      <td>-0.0005</td>\n",
       "      <td>-0.0006</td>\n",
       "      <td>0.0090</td>\n",
       "      <td>0.0060</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>192705</td>\n",
       "      <td>5</td>\n",
       "      <td>0.022322</td>\n",
       "      <td>1</td>\n",
       "      <td>0.059578</td>\n",
       "      <td>-3.025963</td>\n",
       "      <td>-2.975058</td>\n",
       "      <td>-2.531446</td>\n",
       "      <td>-0.494518</td>\n",
       "      <td>0.000392</td>\n",
       "      <td>0.00037</td>\n",
       "      <td>0.434783</td>\n",
       "      <td>0.055279</td>\n",
       "      <td>0.0333</td>\n",
       "      <td>0.0327</td>\n",
       "      <td>0.0109</td>\n",
       "      <td>-0.0006</td>\n",
       "      <td>0.0093</td>\n",
       "      <td>-0.0120</td>\n",
       "      <td>0.005780</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   yyyymm  Month   his_avg  bull   premium       d/p       d/y       e/p  \\\n",
       "1  192701      1 -0.005602     0 -0.005602 -2.942374 -2.963349 -2.374773   \n",
       "2  192702      2  0.018589     1  0.042780 -2.979535 -2.932946 -2.430353   \n",
       "3  192703      3  0.013945     1  0.004657 -2.976535 -2.970053 -2.445079   \n",
       "4  192704      4  0.013008     1  0.010196 -2.984225 -2.967143 -2.471309   \n",
       "5  192705      5  0.022322     1  0.059578 -3.025963 -2.975058 -2.531446   \n",
       "\n",
       "        d/e      svar      csp       b/m      ntis     tbl     lty     ltr  \\\n",
       "1 -0.567601  0.000470  0.00037  0.443706  0.050834  0.0323  0.0351  0.0075   \n",
       "2 -0.549182  0.000287  0.00037  0.428501  0.051682  0.0329  0.0347  0.0088   \n",
       "3 -0.531456  0.000924  0.00037  0.469765  0.046370  0.0320  0.0331  0.0253   \n",
       "4 -0.512916  0.000603  0.00037  0.456754  0.050518  0.0339  0.0333 -0.0005   \n",
       "5 -0.494518  0.000392  0.00037  0.434783  0.055279  0.0333  0.0327  0.0109   \n",
       "\n",
       "      tms     dfy     dfr      infl  \n",
       "1  0.0028  0.0095 -0.0019 -0.011299  \n",
       "2  0.0018  0.0092 -0.0019 -0.005714  \n",
       "3  0.0011  0.0092 -0.0170 -0.005747  \n",
       "4 -0.0006  0.0090  0.0060  0.000000  \n",
       "5 -0.0006  0.0093 -0.0120  0.005780  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# reorganize the dataframe\n",
    "df = df[['yyyymm','Month','his_avg','bull','premium','d/p','d/y','e/p','d/e','svar',\n",
    "          'csp','b/m','ntis','tbl','lty','ltr','tms','dfy','dfr','infl']]\n",
    "df.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plot the heatmap of feature correlation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1a209e1518>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 792x648 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "features = df[['premium','d/p','d/y','e/p','d/e','svar',\n",
    "          'csp','b/m','ntis','tbl','lty','ltr','tms','dfy','dfr','infl']]\n",
    "\n",
    "corr = features.corr()\n",
    "\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Generate a mask for the upper triangle\n",
    "mask = np.zeros_like(corr, dtype=np.bool)\n",
    "mask[np.triu_indices_from(mask)] = True\n",
    "\n",
    "# Set up the matplotlib figure\n",
    "f, ax = plt.subplots(figsize=(11, 9))\n",
    "\n",
    "# Generate a custom diverging colormap\n",
    "cmap = sns.diverging_palette(220, 10, as_cmap=True)\n",
    "\n",
    "# Draw the heatmap with the mask and correct aspect ratio\n",
    "sns.heatmap(corr, mask=mask, cmap=cmap, vmax=.3, center=0,\n",
    "            square=True, linewidths=.5, cbar_kws={\"shrink\": .5})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We recognize the autocorrelation in time series. In order to take full advantage of the past information, we include the past 20 years' data in the prediction. Here we generate predictors for the past 20 years, i.e. 240 months"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "past = 241\n",
    "\n",
    "num = len(df.columns)\n",
    "for i in range(4,num):\n",
    "    name = df.columns[i]\n",
    "    for j in range(1,past):\n",
    "        t = str(j)\n",
    "        df[name+\"-\"+t] = df[name].shift(j)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# drop the first past-20 years predictors\n",
    "df = df[past-1:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# drop the contemporanous predictors\n",
    "df = df.drop(columns=['d/p','d/y','e/p','d/e','svar',\n",
    "          'csp','b/m','ntis','tbl','lty','ltr','tms','dfy','dfr','infl'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3845"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# column number 3845: 'yyyymm','Month',‘his_avg’，'bull','premium' + 16 indicator * 240 months \n",
    "len(df.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>yyyymm</th>\n",
       "      <th>Month</th>\n",
       "      <th>his_avg</th>\n",
       "      <th>bull</th>\n",
       "      <th>premium</th>\n",
       "      <th>premium-1</th>\n",
       "      <th>premium-2</th>\n",
       "      <th>premium-3</th>\n",
       "      <th>premium-4</th>\n",
       "      <th>premium-5</th>\n",
       "      <th>...</th>\n",
       "      <th>infl-231</th>\n",
       "      <th>infl-232</th>\n",
       "      <th>infl-233</th>\n",
       "      <th>infl-234</th>\n",
       "      <th>infl-235</th>\n",
       "      <th>infl-236</th>\n",
       "      <th>infl-237</th>\n",
       "      <th>infl-238</th>\n",
       "      <th>infl-239</th>\n",
       "      <th>infl-240</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>241</th>\n",
       "      <td>194701</td>\n",
       "      <td>1</td>\n",
       "      <td>0.007164</td>\n",
       "      <td>1</td>\n",
       "      <td>0.021696</td>\n",
       "      <td>0.046501</td>\n",
       "      <td>-0.004118</td>\n",
       "      <td>-0.005953</td>\n",
       "      <td>-0.104430</td>\n",
       "      <td>-0.069172</td>\n",
       "      <td>...</td>\n",
       "      <td>0.005780</td>\n",
       "      <td>0.005814</td>\n",
       "      <td>-0.005780</td>\n",
       "      <td>-0.017045</td>\n",
       "      <td>0.011494</td>\n",
       "      <td>0.005780</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.005747</td>\n",
       "      <td>-0.005714</td>\n",
       "      <td>-0.011299</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>242</th>\n",
       "      <td>194702</td>\n",
       "      <td>2</td>\n",
       "      <td>0.007099</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.008624</td>\n",
       "      <td>0.021696</td>\n",
       "      <td>0.046501</td>\n",
       "      <td>-0.004118</td>\n",
       "      <td>-0.005953</td>\n",
       "      <td>-0.104430</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.005747</td>\n",
       "      <td>0.005780</td>\n",
       "      <td>0.005814</td>\n",
       "      <td>-0.005780</td>\n",
       "      <td>-0.017045</td>\n",
       "      <td>0.011494</td>\n",
       "      <td>0.005780</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.005747</td>\n",
       "      <td>-0.005714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>243</th>\n",
       "      <td>194703</td>\n",
       "      <td>3</td>\n",
       "      <td>0.007006</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.015325</td>\n",
       "      <td>-0.008624</td>\n",
       "      <td>0.021696</td>\n",
       "      <td>0.046501</td>\n",
       "      <td>-0.004118</td>\n",
       "      <td>-0.005953</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.005747</td>\n",
       "      <td>0.005780</td>\n",
       "      <td>0.005814</td>\n",
       "      <td>-0.005780</td>\n",
       "      <td>-0.017045</td>\n",
       "      <td>0.011494</td>\n",
       "      <td>0.005780</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.005747</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>244</th>\n",
       "      <td>194704</td>\n",
       "      <td>4</td>\n",
       "      <td>0.006814</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.039877</td>\n",
       "      <td>-0.015325</td>\n",
       "      <td>-0.008624</td>\n",
       "      <td>0.021696</td>\n",
       "      <td>0.046501</td>\n",
       "      <td>-0.004118</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.005747</td>\n",
       "      <td>0.005780</td>\n",
       "      <td>0.005814</td>\n",
       "      <td>-0.005780</td>\n",
       "      <td>-0.017045</td>\n",
       "      <td>0.011494</td>\n",
       "      <td>0.005780</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>245</th>\n",
       "      <td>194705</td>\n",
       "      <td>5</td>\n",
       "      <td>0.006781</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.001307</td>\n",
       "      <td>-0.039877</td>\n",
       "      <td>-0.015325</td>\n",
       "      <td>-0.008624</td>\n",
       "      <td>0.021696</td>\n",
       "      <td>0.046501</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.011561</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.005747</td>\n",
       "      <td>0.005780</td>\n",
       "      <td>0.005814</td>\n",
       "      <td>-0.005780</td>\n",
       "      <td>-0.017045</td>\n",
       "      <td>0.011494</td>\n",
       "      <td>0.005780</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 3845 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     yyyymm  Month   his_avg  bull   premium  premium-1  premium-2  premium-3  \\\n",
       "241  194701      1  0.007164     1  0.021696   0.046501  -0.004118  -0.005953   \n",
       "242  194702      2  0.007099     0 -0.008624   0.021696   0.046501  -0.004118   \n",
       "243  194703      3  0.007006     0 -0.015325  -0.008624   0.021696   0.046501   \n",
       "244  194704      4  0.006814     0 -0.039877  -0.015325  -0.008624   0.021696   \n",
       "245  194705      5  0.006781     0 -0.001307  -0.039877  -0.015325  -0.008624   \n",
       "\n",
       "     premium-4  premium-5    ...     infl-231  infl-232  infl-233  infl-234  \\\n",
       "241  -0.104430  -0.069172    ...     0.005780  0.005814 -0.005780 -0.017045   \n",
       "242  -0.005953  -0.104430    ...    -0.005747  0.005780  0.005814 -0.005780   \n",
       "243  -0.004118  -0.005953    ...     0.000000 -0.005747  0.005780  0.005814   \n",
       "244   0.046501  -0.004118    ...     0.000000  0.000000 -0.005747  0.005780   \n",
       "245   0.021696   0.046501    ...    -0.011561  0.000000  0.000000 -0.005747   \n",
       "\n",
       "     infl-235  infl-236  infl-237  infl-238  infl-239  infl-240  \n",
       "241  0.011494  0.005780  0.000000 -0.005747 -0.005714 -0.011299  \n",
       "242 -0.017045  0.011494  0.005780  0.000000 -0.005747 -0.005714  \n",
       "243 -0.005780 -0.017045  0.011494  0.005780  0.000000 -0.005747  \n",
       "244  0.005814 -0.005780 -0.017045  0.011494  0.005780  0.000000  \n",
       "245  0.005780  0.005814 -0.005780 -0.017045  0.011494  0.005780  \n",
       "\n",
       "[5 rows x 3845 columns]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# now the features include the past 20 years observations\n",
    "df.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we plot the distribution of the equity premium. The returns are left-skewed (i.e. negative skewness)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "hist = df[\"premium\"].hist(bins=10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classification (Bull Market Prediction)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For classification purpose, we first normalize the features "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "bull = df.bull\n",
    "x = df.iloc[:,5:]\n",
    "norm_x = (x- x.mean())/x.std(ddof=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We use the first half as training data and the second half as testing data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "norm_x_train = norm_x[0:int(0.8*len(norm_x))]\n",
    "norm_x_test = norm_x[int(0.8*len(norm_x)):]\n",
    "bull_train = bull[0:int(0.8*len(norm_x))]\n",
    "bull_test = bull[int(0.8*len(norm_x)):]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Decision tree (CART algorithm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix:  \n",
      " [[18 37]\n",
      " [44 62]]\n",
      "precision:  0.6263\n",
      "recall:  0.5849\n"
     ]
    }
   ],
   "source": [
    "from sklearn import tree\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "clf = tree.DecisionTreeClassifier(max_depth=50)\n",
    "clf = clf.fit(norm_x_train, bull_train)\n",
    "predicted = clf.predict(norm_x_test)\n",
    "c_matrix = confusion_matrix(predicted,bull_test)\n",
    "\n",
    "# precision = True Positive/(True Positive + False Positive)\n",
    "precision = c_matrix[1,1]/(c_matrix[1,1]+c_matrix[0,1])\n",
    "# Recall = True Positive/(True Positive + False Negative)\n",
    "recall= c_matrix[1,1]/(c_matrix[1,1]+c_matrix[1,0])\n",
    "\n",
    "print(\"Confusion Matrix: \" ,\"\\n\" , c_matrix)  \n",
    "\n",
    "print(\"precision: \", round(precision,4))  \n",
    "print(\"recall: \", round(recall,4))  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "temp = df['premium']\n",
    "premium = temp[int(2*len(temp)/3):]\n",
    "cumsum = 0\n",
    "cs = []\n",
    "\n",
    "cumsum_h = 0\n",
    "cs_h = []\n",
    "\n",
    "for i in range(int(len(predicted))):\n",
    "    if predicted[i] == 1:\n",
    "        cumsum += premium[i+884]\n",
    "    else:\n",
    "        cumsum += -1*premium[i+884]\n",
    "        # unblock the above two lines to impose short-selling constraint\n",
    "    cs.append(cumsum)\n",
    "\n",
    "for i in range(len(bull_test)):\n",
    "    cumsum_h += premium[i+884]\n",
    "    cs_h.append(cumsum_h)\n",
    "    \n",
    "plt.figure()\n",
    "plt.plot(cs, label='ML Portfolio Return')\n",
    "plt.plot(cs_h, label='Benchmark Return')\n",
    "plt.legend(loc='best')\n",
    "plt.ylabel('Return ($)')\n",
    "plt.xlabel('Months')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Naive Bayes "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix: \n",
      " [[48 66]\n",
      " [14 33]]\n",
      "precision:  0.3333\n",
      "recall:  0.7021\n"
     ]
    }
   ],
   "source": [
    "from sklearn.naive_bayes import GaussianNB\n",
    "\n",
    "gnb = GaussianNB()\n",
    "\n",
    "gnb.fit(norm_x_train,bull_train).predict(norm_x_test)\n",
    "\n",
    "predicted= gnb.predict(norm_x_test)\n",
    "\n",
    "c_matrix = confusion_matrix(predicted,bull_test)\n",
    "precision = c_matrix[1,1]/(c_matrix[1,1]+c_matrix[0,1])\n",
    "recall= c_matrix[1,1]/(c_matrix[1,1]+c_matrix[1,0])\n",
    "\n",
    "print(\"Confusion Matrix: \"\"\\n\" , c_matrix)  \n",
    "\n",
    "print(\"precision: \", round(precision,4))  \n",
    "print(\"recall: \", round(recall,4)) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "temp = df['premium']\n",
    "premium = temp[int(len(temp)/2):]\n",
    "\n",
    "cumsum = 0\n",
    "cs = []\n",
    "\n",
    "cumsum_h = 0\n",
    "cs_h = []\n",
    "\n",
    "for i in range(len(predicted)):\n",
    "    if predicted[i] == 1:\n",
    "        cumsum += premium[i+884]\n",
    "    else:\n",
    "        cumsum += -1*premium[i+884]\n",
    "        # unblock the above two lines to impose short-selling constraint\n",
    "    cs.append(cumsum)\n",
    "\n",
    "for i in range(len(bull_test)):\n",
    "    cumsum_h += premium[i+884]\n",
    "    cs_h.append(cumsum_h)\n",
    "    \n",
    "plt.figure()\n",
    "plt.plot(cs, label='ML Portfolio Return')\n",
    "plt.plot(cs_h, label='Benchmark Return')\n",
    "plt.legend(loc='best')\n",
    "plt.ylabel('Return ($)')\n",
    "plt.xlabel('Months')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "K-Means (K=7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix: \n",
      " [[55 84]\n",
      " [ 7 15]]\n",
      "precision:  0.1515\n",
      "recall:  0.6818\n"
     ]
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "clf = KNeighborsClassifier(n_neighbors=7)\n",
    "clf.fit(norm_x_train, bull_train)\n",
    "predicted= clf.predict(norm_x_test)\n",
    "\n",
    "c_matrix = confusion_matrix(predicted,bull_test)\n",
    "precision = c_matrix[1,1]/(c_matrix[1,1]+c_matrix[0,1])\n",
    "recall= c_matrix[1,1]/(c_matrix[1,1]+c_matrix[1,0])\n",
    "\n",
    "print(\"Confusion Matrix: \"\"\\n\" , c_matrix)    \n",
    "\n",
    "print(\"precision: \", round(precision,4))  \n",
    "print(\"recall: \", round(recall,4))  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY0AAAEKCAYAAADuEgmxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzsnXd4W9XZwH9H8t577+zE2c4gYQRIgBAg7L1nBy2jpaUf/ejXQVtaSkvLaGmhhLLKCgSSACGEACF7x1l2HO+999T5/jiSp7xlSbbP73n0SLr36N43jnTf+24hpUSj0Wg0moFgcLQAGo1Goxk9aKWh0Wg0mgGjlYZGo9FoBoxWGhqNRqMZMFppaDQajWbAaKWh0Wg0mgGjlYZGo9FoBoxWGhqNRqMZMFppaDQajWbAuDhaAFsTEhIiExISHC2GRqPRjCr27t1bKqUM7W/dmFMaCQkJ7Nmzx9FiaDQazahCCJE1kHXaPaXRaDSaAaOVhkaj0WgGjFYaGo1GoxkwWmloNBqNZsBopaHRaDSaAaOVhkaj0WgGjFYaGo1GoxkwWmloNJoufHQwn7LaJkeLoXFStNLQaDTtlNY28YM39/P6zmxHi6JxUrTS0Gg07eRXNgCQWVbnYEk0zopDlYYQ4iIhxAkhRLoQ4tFe1lwrhDgqhEgVQrxhbxk1mvGERWlkl9U7WBKNs+Kw3lNCCCPwHLACyAV2CyHWSSmPdlozCfgZsFRKWSGECHOMtBrN+CC/shGArHKtNDTWcaSlsRBIl1JmSCmbgbeA1d3W3AM8J6WsAJBSFttZRo1mXGGxNEpqmqhvbnWwNBpnxJFKIxrI6fQ+17ytM5OByUKIbUKIHUKIi6wdSAhxrxBijxBiT0lJyQiJ69zUN7dql4Jm2BRUNba/ztbWhsYKjlQawso22e29CzAJWAbcAPxLCBHQ40NSviilTJFSpoSG9tsOfkzy4lcZrPrr17SZuv8JNZqBk1fZgL+nKwBZ+iZEYwVHKo1cILbT+xgg38qaD6WULVLK08AJlBLRdON0aR01Ta3t7gWNZigUVDWwKDEIgCydQaWxgiOVxm5gkhAiUQjhBlwPrOu25gPgXAAhRAjKXZVhVylHCRa3gnYpaIZKc6uJ4pompkb6EeDlqi0NjVUcpjSklK3A/cCnwDHgbSllqhDiV0KIy8zLPgXKhBBHgS3AI1LKMsdI7NwUVZuzXvQPXTNEiqobkRKiAzyID/LSNyAaqzh03KuUcgOwodu2xzu9lsDD5oemF6SU7ZZGVrl2KWiGhsW1GenvSVywNwdzKh0skcYZ0RXhY4CK+haaW00AZJXqu0PN0MivUkojKsCT+CAv8iobaGkzOVgqjbOhlcYYoMD8YzcaRJ9FWa1tJp785DhfntDlLpqeWAr7ogI8iAv2os0kdWKFpgdaaYwBCs2uqeRof7LL6lBevZ785fM0XvjyFHe8sps/bzqJSafnajqRX9lAgJcrXm4uxAd5AZBpxxhZY0sbVfUtdjufZmhopTEGsMQzFicFUdfcRlldc48136SV8tyX6Vw5N5or58bwzOY0Pj5cYG9RNU5MQVUjUf6eAMQFK6Vhz2D4E+uPceUL2+x2Ps3Q0EpjDFBU3YjRIEiJt+TXd/2hN7eaeOTdg0wI9eE3VyTzh6tn4WY0cDS/2hHiapyU/MoGogI8AAj39cDVKMitsJ/S2JZeyqmSOqoatLXhzGilMQYoqGokzNedxBBvALK7ZVCtP5xPQVUjj62ahpebC0aDIC7Yi9OltY4QV+Ok5Fc2EGm2NAwGQXSAJ7nl9olpVNQ1k1GqvrfpxTV2OadmaGilMQYorGokwt+D2CBPhOhqaUgp+fe2TJJCvTlnUkeLlYRgbzJ1ppXGTEVdM9WNrcSb3VIAsUFedrM09udUtL8+UahuZoprGqms7+lq1TgWrTTGAAVVDUT6e+DuYiTSz6OL0tibVcGh3CruWJKAwdDR7isp1JvMsjodDNcAtN/lJwR7t2+LCfQip8I+lsberAqMBoGnq5GTRcrSuPWlXTy29ohdzq8ZOFppjAEKqxqJ8OsIYHbuGfTKt5n4ebhw5byYLp9JCPamqdVEQXUjGk2mWWkkhnZWGp6U1zVT1zTyLdL3ZVUyLdKXKRG+nCyqoaCqgeOFNe0KROM8aKUxyqlpbKGuuY0If3cAEkN8SC+uRUqJlJKv00pZmRyJt3vX4n9L/ON0ia4g16jxrgYBsYFd3VMAuSNsbbS2mTiYW8n8uECmhCulsS1ddQvKLq/vNYVc4xi00hjlWGo0IswBzJnR/lQ3tpJdXk9OeQNVDS3MivXv8bl2paE7mWpQXZJjAr1wc+m4JMQGqu9Uzgin3Z4oqqG+uY158YFMCvehtLaZjw6qhtdNrSZKappG9PyaweHQ3lOaoVNS08RfPj9JiI+yMCL9VarkrBilIA7mVmEUKoYxK7rHCBLC/dzxdDVqS0MDKEsjIcS7y7aYQIulYTulUdfU2sPq3ZelguDz4gLJNN/EbD1ZQpC3G+V1zWSX1xPm52EzGTTDQ1sao5RPjhTw+s5sntmcBkCE+Uc1JcIXdxcDh3IqOZxXhatRMDnCp8fnhRAkhHi3/0g14xcpJZml9SR2ypwCCPFxw8PVYLNg+P7sCmb/8jP2Z1d02X6ssAZ/T1diAj2ZEu7bvv3q+SoOp7vtOhdaaYxS0otr8XF34deXJ3Pn0kSiA5QrwdVoYHqUH4dyqzicV2lWIkarx0gK8eZ0qVYa453S2mZqm1p7WBpCCGICbZd2+5/tWbSaJDsyyrtsP11SR1KoN0IIQn3d2ycHXpsSixBaaTgb2j01SkkrrmVCmA+3LI7vsW92TAD/3Z2Dq1GwalZUr8dICPHi09RCWtpMuBr1/cN4xXLjkNhNaYCKa+TYoMCvqr6F9ea2NUfyq3qcf+nEEEApqqkRvuRXNTAxzIdIPw+tNJwMfaUYpaQX1zIprKfbCVRco6GljerGVmZG9wyCW0gM8aHVJEc8O0bj3GT2pTSCvMixgaXxwYE8mlpNJIV6cySvQ2nUNbVSWN1IUqdU319fnszzN87vOL9WGgOj5CTUl/e/bphopTEKqWpoobimiYm9Ko2OwHffSkP9UNOLdTuR8czpsjpczG1DuhMT6ElNY+uw+kFJKXlzVzYzo/25al4MWWX17cezxNQ6K6zJ4b7MNCd0xOkJggPnowfg5YtghFOUtdIYhVgu8hNDrSuNpBBvfN1deg2CW5gR5YeLQbCvW2BSM77ILK0jLsgLFysuSkvdxnDu9lPzqzleWMN1C2JJNt/EpJpdVBklvVs5oJRGUXUTjS1tQz7/uCBnF2R/Cyl3gBD9rx8GWmmMQk6ZlcakcOsKwWAQzE8IZFZMQK9BcAAPVyMzov3Zm6WVxniltqmVQ7lVPYLgFizW7PHCoVdmrz9cgNEgWDUzkuQoPwBS81SH5b7iKdDRot2e3XZHJd/8BTwDYd6tI34qrTRGIWnFNbi5GNrz6K3x52vn8OIt8/s9Vkp8IAdzKtvHxWrGD7VNrdz28i4KqxutJlQATAj1wdfdpUea7ECRUrLhcAFLJ4YQ6O1GsI87Uf4e7cHw06V1RAd44uFq/ebGUpWuXVR9UHISTqyHBfeAm3Xla0u00hiFpBfXMiHUB6OhdzPU8gPtj/nxgTS1mtrdBZrxw2NrD3Mgp5Jnb5jLuVPDrK4xGARz4gLYl105pHOk5leTVVbPqpkR7dtmRPtz2BwMzyit69XKAOWeAsi24wTBUceO58DFAxbdZ5fTaaUxCkkrru01CD5YUuIDAbSLahyyM6Ocy2ZHsXJmZJ/r5sYGcKKwekiNCz8+VICLQXDB9A6lMTPan9OlddQ2tXK6pLZPpRHs7YaXm1HXE/VGUy0cfheSrwbvELucUiuNUUZ9cyt5lQ29ptsOljA/NYdDK43xRXVjC4XVjb3GxTozNz4Qk4RDuYOzRk0myfrD+e2uKQuLEoOQEn65LpXqxtY+lYYQgnlxgXydXjqoc48bUtdCc61dYhkWHKo0hBAXCSFOCCHShRCP9rHuaiGEFEKk2FM+ZySjpA4psZmlATA/LpA9WRW6m+g4oj2ZIsy3n5Uwx5zC3XlQ0kB4fWcWOeUNXJPStS3/oqRgbloUxzt7cwG61GhY44IZ4WSU1Nk9NXxbein3vLqH1jYnjvftexVCJkPsQrud0mFKQwhhBJ4DVgLTgRuEENOtrPMFfgjstK+Ezkl7uq0tlUZCECU1TezR1sa4Ia1dafT/PQr0diMpxJt9WQOPaxRWNfLkJyc4a1IIq6y4v/73kunMMGdSJYX0LcPyaeEAbDpaNODz24I3d2Wz6WgRB3Ot/7tLapqotcOskV4pPg65u5SVMcJptp1xpKWxEEiXUmZIKZuBt4DVVtb9GvgDoKcFoZSG0SC6TFgbLhcnRxAb5Mmdr+zmYM7QAp6a0UV6cS1uLob27KT+mBsXyIGcgVujv/wolVaTiScun4mwckHzcDXyr9tS+PXlycQG9Swq7ExUgCczo/357GjhgM5tC0wmyben1EyPr05ad41d9+J2Hlt72G4y9eDQf8HgArOut+tpHak0ooGcTu9zzdvaEULMBWKllB/3dSAhxL1CiD1CiD0lJSW2l9SJSCuuIT6469yD4RLs485b955BgJcrt7y0U7dtGAekF9eSFOLdZwZeZ+bFB1Ba28x5f9rKY2sP9zkmuKG5jU9TC7n1jIT2OgtrRPp7csvieKtKpTsrpodzIKeSYjtNmjxWWE15XTMGAV+l9bymlNU2kVFSxxfHimlxlPsqYwvELACfULue1pFKw9o3pf2bKIQwAH8GftTfgaSUL0opU6SUKaGh9v0D2pu+ek4Nh+gAT964ezFSwsNvH6BNzw4f06QV1zApvP94hoXVc6J55MIphPi48frO7D77UR0tqMYkOzLzbMEFM8KREq76+7dc/ty2Lv2rRoJt5sD7tSmxHMyppKq+axsVS8pwTVOrY5JIGiqg4CAknmP3UztSaeQCsZ3exwD5nd77AsnAl0KITGAxsG48B8ObW01kltXbNJ7RmdggL365ega7Myt48auMETmHxvE0NLeRW9HQaxsaa/i4u/D9cyfy6MppAKQV9R6UtlzQLf2jbMGUcF/uOzuJqRF+HMqt5LNU27uqWtpMFNcoS+ab9DImhflwTUoMJgnfdMvesvwbXQyCLSeKbS5Lv2R9C9IEiWfb/dSOVBq7gUlCiEQhhBtwPbDOslNKWSWlDJFSJkgpE4AdwGVSyj2OEdfxZJXV0WaSA8p4GSpXzI3m4pkRPL3pBGlFqnWElFJbHmOIUyW1SNl7G5q+sNywpJf0rjQO51UR4uPWPhjMFggh+NnF0/jnrSkkhnhzbBhtTbpTZ66Mn/V/n7Hwic08/uERdp0uY+nEEGbHBODr4cJXJ7u6qA7nVZEQ7MWipCC2HHeA0sjYCi6eyj1lZxymNKSUrcD9wKfAMeBtKWWqEOJXQojLHCWXM5M2AplT3RFC8OvVyfi4u/Do+4cpq23iiue/5Z5Xx62uHnOkDyJzqjv+nq6E+br3mf56JK+K5Gj/AcUqhsLUSD9O2FBppOZXs/VkCcunh3Pjojhe3Z5FY4uJMyeG4GI0cObEELaeLOkSxzmSV83MmADOnRLGyaJa+/fGOv0VxJ8BLm79r7UxDq3TkFJukFJOllJOkFI+Yd72uJRynZW1y8azlQEdP/b+8tqHS7CPOz9fNZ29WRWc//RWDuRU8nVaie40OkZIK67BxSCIH2IG3sQwn16VRmNLG2nFtX225B8u0yJ8yS6vt1m6a0GVmifzw/Mm8tsrZvLsjXO5eGYESyYGA3BRcgSF1Y3sPK1mVZTVNpFX2cDMaD+WTVHtVz5LtWM6cG0xlBxziGsKdEW4Q5BSUlbbNOjPpRXXEhPoiZfbyA9cvHJeNGdNCqGpxcQdSxNoaZOk5leP+Hk1I09mWT2xQUPPwLMoDWvpt0cLqmkzyfYW6CPB1AhV32Era6OwSsUxIvyVO+2SWVE8f9P89t/ZBdMj8HYzsna/Kka0BMGTo/2ZEOrN3LgAnvrsBEdH6vdhaoNDb0Nrs3qfsVU9OyAIDlppOITfrD/GmU9uGfSdUroNe071hxCCf96awlc/OZfvnjMBYMidTjXORUVdM8HeQ3drTAzzobaplaLqnjc+7UHwkVQakSqmd7zQNhfpgqpGfNxd8PVwtbrf083IypmRbDhcSENzW/u/0eKCe+Gm+fh6uHD3mt2U1Az+ZrBf0j6D9++BfWvU+8Nvg28kRM62/bkGgFYadmZHRhkvfXOahpY2MvoIJnanzSTJKBmZdNve8HA1EurrTpifB9EBnuzXhX9jgsr6FgK8rF8gB0J7MNyKi+pwbhXB3m5E+tsuCN6d6ABPfN1dOF5gO0ujP3mvnBdNbVMrr+3IYtPRIhJDvPEzK5kIfw9eum0BpXXN/H7jcZvI1IWsbep590tQXQDpn8Ps68HQ+6yckUQrDTtS19TKI+8ebP/Bnhqg0mgzSZ767ARNraZ209zezI0L4MAQ22NrnIvK+mYCvIZnaYCKjXSmsaWNb0+VMTNm5ILgoKzgKRG+NrQ0GtpdU72xODGYKH8PnthwjKMF1dy5NKHL/uRof25fksD7+3M5WWS7ID0A2TtU5XfJMfj4IZVqO+dm255jEGilYUc2Hikkp7yBZ66fi9EgOFXcf7tnKSV3r9nNC1+e4rqUWC6dHWUHSXsyNy6QvMoGiuxUkasZOSobWgjwHLqlEerjjp+HSw9L46+b08irbOCOpYnDFbFfpkb6crywxiZNNgsGYGkYDILHVk3njqUJbPnxMm45I6HHmu+eMwEfNxee+vTEsGVqp7ke8vdDyl3g7g8nN0LsYgiZaLtzDBKtNOzIqZJaXAyCpROCiQvyIqO0f0sjq6yeLSdK+OF5E3ny6lk2bR8yGObGmTudamtjVNPU2kZ9c9uw3FNCCCaF+3ZRGkfyqvjHVxlcMz+GcyaPfFeGqRF+1DS2kl81vJuYljYTJbVNRPj33f8KYNWsSH5x6YxeJ2YGertxz9lJfHa0yHbxv7y9YGqFiefDnBvUtrk32ebYQ0QrDTuSWVpHXJAXLkYDE0K9B2RpWMZirug0xMYRzIjyw81oYJ8Oho9qLO0whuOeApgY6kNapwyq3244RpC3Gz9f1aNR9Ygw3dwh99thztkormlCSmwWg7nrzEQCvVx5bks6oH7zP333EFtPlgzcKpISdr4IJScgezsgVOvzpQ/Awvsg+SqbyDpUtNKwI6dL60gwD5yZEOrDaXOFd1+k5lfjYhBMjrBfANwa7i5GFiUF8fI3p3n2izRdIT5KqWywKI2hWxoAi5KCKK9rZl92JSU1TWzPKOOGhXH4D/O4A2VOTABTI3x5bkv6sBoGFpprNPqLaQwUb3cX7liayOfHijmSV8UD/z3Af/fkcNvLu1j5zNccHsggqxMbYOMjsOYyOP4xhE0Hz0Dwi4KL/2CXOeB9oZWGnZBSklVW397SPCnUm+ZWE3kVDT3Wbj5WxCFzD//U/Gomhfvi7uKYTInOPHvDPC5KjuCpz07y9CYb+m01dqPSYml4Ds/SWDE9HHcXAx8dzOeT1EKkxOrcjJHCYBD86IIpZJbV8/6+3CEfJ79SubeiBuCeGii3nZGAj7sLt728i4M5lTx97WyevnY2VQ0tXPH8Nv697XTvHzaZ4IsnwD8OWhpUU8L4M2wmmy3QSsNOFFU30dDSRmKI8odOMDeL655Bdbq0ju+8tpf//eAIUkqO5le1D6txNP5ervzthrksmRDMF8fHdgv6sUpFvSoQG66l4evhynlTw/j4UAEfH8wnKdSbyUPoZTUclk8LY3ZsAH/dnE5T69C6FXQv7LMF/l6u3Lw4nrK6ZlYmR3DF3GiunBfDJw+czZKJIfz646O912ilvg/FqbD8F3D9a+DqBZMuAFRdynX/2D6kwmBbopWGnbAEvRPNU8p6UxpPrD9GS5vkYG4V+7IrKK1tdhqlASoIuigxmOOF1VQ1tPT/AY1T0RHTGL4b6dLZUZTWNrHzdDmrZkaOaJqtNYQQPLh8EnmVDUNu41FQ1YiXmxE/D9t2WfjuORO4/9yJPHFFxxAqfy9X7jozEZPEevp6axNseQLCZsCMK1WbkEdzYPKFAPxjawY7T5fz3jAsK1uglYadyCxVDc0SzJZGoLcbQd5unCrpCIZ/k1bK58eKuGFhHAB/+ES5gGZEjVx17VBYkBCIlOig+CikssFiaQy/0d15U8PwcVcX25XJ9nNNdeacSaFE+XsM+UJaWK1qNGyt8Py9XPnxhVMI6lZ5PzcuACGwPoPj279CeQas+BUYzJdmo/r7ltc1s/5QAQDv7Mm1SarxUNFKw05kltXh5mLo4jtNCvFutzSqG1v4+QeHiQ3y5BeXTmdmtH97g7TpTmRpAMyJC8BoEOzJLHe0KJpBUlHfgqtR4O02/BiZh6uR1XOimB7px7TIkWvX3xcGg+DyudF8dbJkSFP9BlKjYUv8PFyZEu7Lnqyuv53mkgzkV0/B9NUwaXmPz72zJ4fmNhN3Lk0krbiWgwMJqI8QWmnYidOldcQHeWHoNF5zQqgPR/Or2X6qjJ++e4icigb+dM0cPFyNXGwOKiaGeLffzTkLXm4uJEf5sTtTWxqjjcr6Fvw93Wx2Z/2r1cl88P2ldndNdeaq+WpQ0ocH8vtf3I2CykYi/GwXBB8I8+MDOZBd2SUDce8/v0dTm0Be+Lse600myRu7slmYGMRDKybh4WrgnT05PdbZC6007ERmp3RbC3eemUiQtxs3/HMHG48U8tOLprAwMQjoyERxNivDQkpCEAdyKoccgNQ4hqqGZpvEMywYDcJhBacWJoT6MCc2gLd2Z7P5WBHpxQNr49FqntQXFWA/SwOU0qhpam1vN5KRlc3Cph283LKCT3N6/i33ZFWQVVbPTYvi8PVwZWVyJOsO5jtsVIFWGnagzaTSbZO6KY0pEb58+uDZfP/cCdy5NJF7zkpq3xcX7MWPL5jMbVbaFTgDCxICaW41DXpWc3pxLR/szzNPj9O1Hvamoq6FQDvVUtiTGxbGcqqkjrvW7OHiZ75pH9vaFwVVjZgkRAXY19JIiVc3hnvMcY3Mbe9gFJL9Pufwy4+OUtcts2pnRhlAe6X9xTMjqWlsdVh3Bq007EB+ZQPNbaYelgaotsuPXDiVxy+d3sPEv/+8Se2Wh7ORkqDk2tJL6u2HB/I4+w9b2utNLPzonYM8+N8DnP+nrVz+/LfkVfasU9GMHJUNyj011rg2JZYtP17Gv29fQHObibd39+++sdzp2ztVODbIk1Bfd/aZlYbP6U8oNoTxneuvoKCqkWfN1eQWdmdVMDncpz15YVFSEAYB20+pavjCqsZB37wNB6007EBGqcqQSrSiNEYrIT7urEyO4MWvM3qkDacV1fDoe4fJLq/ntpd3tc8azyyt42BOJfeclcgvL5tBRnEtl/7tG3ad1gF1e1FVb1v3lLMghCAxxJtzp4axZEIwb+7Koc0kOV1a1z4cqc2kmn+u+TYTgBPm7+XEMPsG8YUQLEwM4vNjRXx1JIPZzfsoiFrB/IRgLp8Txb+3nW63lNpMkn1ZFSxI6Lh59PNwZVZMANtOKQvkkXcPct0/trenU480WmnYgRPmFs5Twh2TYTJS/HL1DDxdjfz03UPt85Mbmtv4/hv78HIz8vZ9Z+BiNHDLS7uorG9m3cF8hFCxnNuWJPDh/UsJ8HLlzld2k5rvuGyQ8URF/dh0T3Xm5sXx5FU28KfPTnDp377hmr9/y+nSOl7fmcXnx4pZuz8PgLSiWiL9PfAfRsffofLoRVPxdDXyzpsv4y5aCV14NQAPLp9MS5vk+S2nADhWUE1tU2sPj8OSCcEczKnkeGE1X6eVUtfcxms7s+wiu1YaduBYQQ0Rfh4EDmNamjMS5uvB45dMZ09WRbtJ/X/rUkkrruXP181hYWIQL9+2gJLaJn798TE+OJDHosQgIs1px0mhPrx+9yJ8PVy4/d+7ya2od+Q/Z8zT2NJGQ0ubTWo0nJkV08MJ9XXn+S9PEebnjovRwPdf38cfPz2BQcDR/GqaWts4UVjDZAfdyMUGefGfO1K40fVLKkQAUclqdGtCiDfXpsTwxs5scivq2W1Oa09J6K40Qmg1SX763mGEUJMS/73ttF2C41pp2IFjBdXtIyrHGlfOi+aKudE8vekkD5mbs31v2QTONgftZsb4851zknhvXy4ZJXWsnhPd5fOR/p6suXMhtY2t/G1zurVTaGxEtbmC3xF31vbE1Wjg4RWTOWdyKO/cdwZPXjWLowXVNLa08aMLptDcZuJQbhXpJbVMiXDQ71JKpuz7JWdwGMPZP+oyhe8H501CCHjwrQN8k1ZKdIAn0d2C9SkJgbgZDRzMqeScyaH87OKplNY226VaXCuNYfLkJ8f586aTve5vbjVxqqSWaZHOmTo7XIQQPHnVLJZMCGbt/jwWJgTx0PLJXdb84LxJTAj1xtUoWJncs8X75HBfzp8WxubjRbp77ghSYfZ5B45xSwPghoVxrLlzIcE+7lyUHMHPV03jt1fM5Jr5MQCs3Z9Hc6vJruOTu7DjedjzMix9EP9zf9hlV1SAJ09fO4c9WRVsPl7MgoTAHh/3cDUyL17NuLkuJZYzkoKZHePPv7dljnhWokOrxoQQFwHPAEbgX1LK33fb/zBwN9AKlAB3Sint47gbIO/tzcXH3YWHVky2uv9USS0tbZKpjrqjsQNuLgb+fst8Xvr6NDctisPF2PVexMPVyEu3LSCzrK5X18iK6eF8fKiAAzkVzI93zoyx0U6ljZoVjkbu7pTOHh3gyTpzIaDDLI1Db0PMQlj+f1Z3r5oVSUHVNH6z/hiLk4Ktrlk9J5rK+hbOnxaOEILfXjmTQC/bFW72hsMsDSGEEXgOWAlMB24QQnSf4LIfSJFSzgLeBf5gXyn7pryumeKaJrLK62lutd7T3zLHeKxaGhb8PFx5aMVkwvysF0olhHizbEpYr59fNiUMF4Pgs6P9N577Nr2Up/uw7jTWqRwn7qn+mBsXQG1TK0J0zDu3K811UHhYNSTs4wJ/15mJrP3u5mM8AAAgAElEQVTeEq6cF2N1/w0L4/jkwbPbiytnRPnbpebEke6phUC6lDJDStkMvAWs7rxASrlFSmmJju4ArP/1HIRFIbSZJJll1qfwHS+owc1oGFPptiOBv6cri5OC2dSP0mhobuNH7xzkr5vTdMbVILFYGmMtIWOwzItT7p7YQC+83BzgbMnbC7INYhf1uUwIwdy4QIdX3HfHkdJEA50rcHLN23rjLmDjiEo0SI4XdLQrSCuyPu/7aEE1k8J9cDU613+8M7JiejgZJXU96j4686+vMyioasTFIHhjZ7YdpRv9dAxg0pYG4LDMKXJ2qufYBY45/zBx5JXMml1mNYIjhLgZSAH+2Mv+e4UQe4QQe0pK7Dcc6ERhDX4eLgih2mNY43hhDVMjxrZrylYsnx4OwMbDBe3bLENyAIqrG3lh6ylWJkdw2ZwoPtif1/swG00PKhtUh1svG3S4Hc3MiPInwMu1XXnYneydEDpNjXAdhThSaeQCsZ3exwA92lQKIZYDjwGXSSmtjqySUr4opUyRUqaEhoaOiLDWOF5YTXK0P9EBnqRbuTvOKqujpKbJYW2jRxvRAZ4sTAzi/X15SCnZeLiAxb/bzLt7VRrh7zYep6XNxKMrp3LTonjqmtt4f18upbVNw5oTPV4orGokzNf2syNGG24uBjY/fA73np3U/2JbYzJB7i6IXWj/c9sIRyqN3cAkIUSiEMINuB5Y13mBEGIu8A+Uwih2gIy90maSnCyqZWqEHxPDfHpYGkXVjdz28i583V04f1q4g6QcfVw9P4aM0jr2ZXcUDP5m/VE+PpTP2v153Hf2BOKDvZkXF8DUCF8e/zCVlN98zt1r9jhYcucns6yO+GAvR4vhFAT7uI+8y7jBSkPB0hPQWAVxi0f23COIw5SGlLIVuB/4FDgGvC2lTBVC/EoIcZl52R8BH+AdIcQBIcS6Xg5nd7LL62loaWNqhC8TQ304VVLbXmPQ2NLGTf/aSUlNE6/cuVAHwQfBxTMj8XQ18tjaI6TmV3PPWYnUNbVy/xv7iQvy4v7zJgIqSPjUNbN55MIprEyO4Ku0EoqGMIRnPJFdVq+Vhr1I2wR/SIJD73Tdfvpr9dxPENyZcWidhpRyA7Ch27bHO73uOcLKSbD0k5oa6YtJSppbTeRW1BMf7M1HB/NJL67lX7emMD9+dPotHYWPuwsrZ0bw/r48wv3c+fGFU/B0NfLXL9L55eoZeLh2+OOTo/1JjvbnVEktG48U8vGhAu46M9GB0jsvNY0tlNU1Ex+sb2BGHCnhy9+pDKkNP4bEs8ArBL75M2z9vYpnBDnANWYjdErPEDlWUIMQMCnMl0nm1srpxWpGxMvbMplirnLWDJ5r5qtQ111nJuLuYuShFZP56pFzObeXOo8JoT7MiPLjo4P5SCl5d28uW04UtzdR1EBWmcpcjw/SlsaIk/GlSqs9435obYQ3roO/zoEtv4Hpl8Odn/RZn+HsDMjSEEIYgNlAFNAApEop+6/CGsOkFdcQH+SFp5uRiaEq0J1eXIuXmwvHCqp58qqZ4z7gOFTOmBDMO985g7mxKrtFCEFcP26VS2dH8fuNx/nJu4d4xxw4Twrx5s/XzWF2rIOyZJyIdqWhLY2R5+s/gU8EnP84+MfAJ49Cwlmw6k8w+UJHSzds+lQaQogJwE+B5UAaqpWHBzBZCFGPClKvkVKOu9SV7PL69h+gv5crYb7u/H3rKfw8XQn0cu3RmE8zOBYkDK6ViEVpvLM3lxsXxbEoMYjfbzzOQ/89wIYHzuri1hooTa1tpBXVYjQIogM98fMYvfUNWeWq+LQ/5asZJlnfQubXcMET4OIOi78LM68B7xBHS2Yz+rM0fgO8ANwnu3XBEkKEATcCtwBrRkY85yW3ooHZMR13sH+5fg6v78jmyxPFfP+8iUO6SGmGTnSAJzcvjsPb3YWfXjgVg0EQ5O3GLS/t4tkv0vnxhVMGdbz/W5fK6zuzaGnr+Nr/8rIZ3LYkwcaS24es0npCfNzxcXdoGHN001ilWoD4RVnfLyV8/ktlZaTc2bF9DCkM6EdpSClv6GNfMfAXm0s0CqhpbKGyvoXYTv7hJRNCWDIhBCmldks5iN9cPrPL+7MmhXLVvBj+vvUUl82JGnAFcF5lA2u2Z3L+1HCumKssxtd2ZPH7jcdZMT3c7jOlbUFWuU63HTbv3wtpn8HsG+Dcx8C/mzfh5KeQswNWPQ1uY/dvPehAuBBiqRDiIjGOr4y5FWqudUxgz4vHOP6zOCWPrZqGQQje3NW15Uhrm4lPjhTwr68zeHtPTpeg+dp9uUgJv7h0OqtmRbJqViR/uHoWJil5Yv0xe/8TbEKWTrcdHjWFSmGEJ8OR9+DD73Xd39YCm38FgYkw71bHyGgn+lUaQohXhRAzzK+/AzwL/AB4aYRlc1o6lIb+ETo7Qd5uLJsSyobDBV0Uw283HOc7r+3jN+uP8ZN3D3H7K7spr2tGSsl7+9SEwc6WZGyQF/efO5H1hwv4Nr3UEf+UIdPY0kZBVSPxQToIPmQOvwPSBFe/rJRC7h4wmafkSQnrfwTFqbDiV2AcvbGvgdCn0hBCxKN6PtWYX9+HUhjfBxYLIeKEEOOusZJlLGmsFUtD43xcMjuKouqm9tGZeZUNvLYjiyvnRXPg8RX89oqZ7MgoY+UzX/H8l6c4XVrHVfN7NlS+5+wkogM8+d3G4yM+6MaW5JSr72tCiL7JGTIH34Lo+RAyCSLnQHMtlJknTe54AfatgTMfgumX9X2cMUB/lsYywB+4CLgMCACSgHNQg5OWAQkjJp2TklvRgKerkaBx3mJ6tHD+1DA8XA18dEi1NnvmczWL48cXTCHAy40bF8Xx/neX4Ofhyh8/PYGnq5GLZ0b2OI6Hq5GHV0zmcF4VGw4X2vXfMBx0uu0wKTwMRUdg1vXqfdQc9Zx/AJpq4fP/g8kXwXmP93qIsUR/gfA1QogzgGtQCuPvUspXhRDewF1SylftIaSzkVNeT0ygp45fjBK8zf2/Nh4uJDrAi3f35nL7ksQuAe3kaH8++sGZPL8lnRDf3rOMLp8bzT++OsVTn53gghnho6Ll/Tdmd5ou7Bsih98BgwskX6Xeh0wBF08oOADuPtDWBGd8HwzO/12wBQPJv/secCHQLKXcbN4WDDwyYlI5ObkVDV383Rrn57LZUaw/VMCTnxxndow/3z93Qo81Hq5GHr6g79Rco0HwyIVTuefVPXx8KJ8r5jrVXLAevLkrm1e+zeT6BbHjfvjSkEnfDHFngLd57KrRBSJmQv5+aKkHN1+IHb0NCAdLv0rDXLi3sdu2bGBcTMDJKKnlifXH8HAzMjXcl/vPm0huRT0pVoa9a5yXC6aH8+qdC5kU7kOk//BiUedPDSMp1Js132Y5rdKQUvL6zmx+sS6VZVNC+fXlyY4WaXRSU6RcU+d3cz1FzYH9r0NFFkw4F1zGj0LuLxD+kRDiUiFEj3QAIUSSuSPtndY+O1bYdLSIzceLOZBdyZ82neTzY8VUN7ZaTbfVOC9CCM6eHDpshQFgMAhuXRzPgZxKDuZYaX/tYJpbTXzntb38/IMjLJkQzHM3zhsVbjSnJONL9Zx0btftkXOgpQ5q8mHSBXYXy5H09026BzgLOC6E2C2E2CCE+EIIkYFqIbJXSvnyiEvpQDLL6gj0cuXzh8/B39OV329UefqxOt12XHPV/Bi83Yy8uj3L0aL04MsTxXyaWsRDyyez5o6FeOsq8KGTsQU8gyBydtftlmA4wESnbcY9IvSpNKSUhVLKn0gpJ6CC4b8GHgaSpZQrpJQf2kNIR5JZWk9CiDeebkauWxDLqRLVw0fXaIxvfD1cuXJeDB8dyievssHR4nQhs0x9R29fkoDBME6TNb5+Gl5YChWZQz+GlHBqCySdA4ZubYEswfCIWeDXM9NuLDNgm1VKmSml3C6lPCClrB9JoZyJzLI6Es2pircsjsfyG9TuKc3dZyXibjRwz5o91Nl6VrnJBLVDm3d/urSeQC9X/L3GdpFZr7Q0wLZnVCzi5Yug5MTQjlN8DGoLYcJ5PfcZXeC8n8Oynw1P1lGIdnRaaGuBnN1Q3TGmvKFZVdImmCfvxQZ5sXxaOAFergSM1x+kpp34YG/+duNcjhdW88BbB2hu7drs+fcbj3PTv3ZwoHvc4/C7KoDaGyYTvHcn/HWuqgMYJFllde3f2XFJ6lporISVf1BV2/+5AurLB3+cjC3quXs8w8KS+2HqxUOXc5SilYaFulJ4aTkc7fC4WdpJd/4BPnnVLN64e7Gu0dAAsGxKGL+4dAafHyvipn/toLS2CVCzVV786hQ7Msq5/LltvPDlKfWBkhPw3l3w/j3K/WGNLU+oC19zDWRtG7RMWWX1JIznQr49L0PwJFh4L9z0NtQWw0c/7P3v3RuntkDwRAiIHRk5RylaaVjwiwTfSMjb174ps1QpjcROP8BAbzemR427zimaPrhtSQLPXD+Hw3lVrH52G6dKanlmcxoerka2/GgZSyYE89I3p1Xvq33meticnaporDtpm+Drp2D2jeDiAae+GJQsjS1t5Fc1jD+l0daq5m/vXQO5u1VrciEgai6c/79w7CPY+8rAj9faBJnfWHdNjXMGpDTMnW03CSFOCiEyhBCnzRlUY4vo+WpMo5lMS/sF3bNH0w+r50Tzzn1LaGpt4+oXvuXjQ/nctiSBuGAvrkmJobS2iSPZxXDwTZh6ibqYffa/tDVUs3Z/LlX1LepAe18B7zC49BmIX6rudgdBTnk9Uo7DPlO7XoQ1lyiLws0XZl/fse+MH0DiObDpcdWtdiDk7ITWht5dU+OYgVoaLwFPA2cCC1BNDBeMlFAOI3oelJ+ChgpAWRrB3m6jemKbxn7MjPHn7fvOwMvNBR83F+49KwmAZZPDMAjI/vYdqC+DlDtg5R+htpC0jc/y0H8Pcv0/d1BeWqzabydfpYrFJpwLpSegKo/cinqyy/rPP8kcj32mpIT9r6m02Pu+ggcOgFenyY8GA1zyZzWve9MA+0Od2qJahyScOTIyj2IGqjSqpJQbpZTFUsoyy2NEJXME0fPVc/5+AE6XjvOAombQJIX6sP6HZ7L+h2e1t+0I9HZjfnwgURnvgn+sunuNXQCRc/A48SG+Hi5klNTy8r/+Bm3NajwotLtGtn/+Lsuf3srdr+7u9/zWXKpjnsJDqi353FuU4rA2KS94Aix9AA79V7md+uPUFxCzADy0K7o7A1UaW4QQfxRCnCGEmGd5jKhkjiDSXLBjdlFlltWNP9+wZtgEeLn1mMV98UR3ZrccoGbKVe05/81TLyeh6Ti3T5P8/eb5LKr7ghqvWGXxAoRNp84thJIDG3F3MZJWXNtvam9mWR0B4y3d9uBbYHTraCjYG2c+DD7hypXVF/XlUHBQu6Z6YaBKYxHKJfVb4E/mx1PDPbl5AuAJIUS6EOJRK/vdhRD/Ne/fKYRIGO45+8QzQGVd5O2jvrmVouomEsebb1gzIqx0P4xRSL4xLmzfttVNuT6u8djNssgWlhiPsslwpgrgAgjBNuZwkXEPb847hpSS44XVfZ5n3GVOtbXAobdVa/LOLilruHlB4tkqtb43TG2w9UlAKvegpgf99hcQQhiAF6SUb9vyxEIII/AcsALIBXYLIdZJKY92WnYXUCGlnCiEuB54ErjOlnL0IHo+ZGwho1jlx2v3lMYWhBd+SZkI5JXTAaw0b3snTRAhJpOcsw7xxqdIYeTZ8hSWVDUS4e9BSU0TP6u+irWRNUzf+zi/cLmQo/nJzI/v/eJ4urSOBeOpmebxj6G+FObcOLD1MQtU1lpVXseM77pSeGkF+ESAqRVyd6nsq5ixF7a1Bf1aGuYut/ePwLkXAulSygwpZTPwFrC625rVwBrz63eB80d8Nnn0PKgt4rFXP8XNxcDsmIARPZ1mjNHWAh9+H9Zc2jEOtLUZkb6Z4vBz2JlZyamSWspqm/jyZAl50SsRpSehLJ2SS9aQYYrkY/OwqG3ppZThT9WV/0VOX81VLt+QmlfV66mbWs3ptuPlRsfUBl/+HkImD7xpYEyKes7d1bGt6AiUZ0BdMVRmwWXPqsC5rsWyykDdU5uEED8WQsQKIYIsj2GeOxrI6fQ+17zN6hopZStQhZrlMWIUB6jGZA+2/pu1d87SczM0A6elAd66SWXynP4Kjryntmd/C03VRC66HBeD4O3dOTz12QlMJsnkFXfDlIvh5neJnL+KmdH+fHhAKY2v0koI9HJlRnQAImkZftRRmpfe6+k/2J+HlJA4XpTGkfeg5Dic+z89e0P1RvhMVf+Su6djm6ULxI1vw49PwrxbbC/rGGKgSuNO1Fzwr4C95seePj/RP9bUePeSzYGsQQhxrxBijxBiT0nJ0Pr1WPimLobftdzAMrmTGetXQ2Pvd3YaTTvV+fDvi1XK7KqnITxZ3QW3tcKJjWB0J2DGBZw/LYw3dmbz1u4cbl+SQFJ8HNzwZntq5+o5URzOq2L9oQK+SStl6cQQ1XQwYhYAbqWptLaZepz+jZ3ZPPr+YZZMCObCGRF2/ac7hLYW2PJbpQSmdXdQ9IGLm0p4yelkaVTnqWe/7vesGmsMSGlIKROtPJKGee5coHN9fgyQ39saIYQLal55jyYyUsoXpZQpUsqU0NDQYQl1MKeS14yXI698CcrSIGv7sI6nGQcUHoYXl0HpSbj+dVhwl2pkV34KXrkYdv4DJq0AN2+uXxBHTVMrwd5u/HD5pB6HunlxPPPjA3ngrf0U1zRx9iTz9zlsOhIDU+Tp9k7LFoqqG/n5B4c5e1IoL9++AA/XAd51j2ZydkHFaTjr4cGPWY1doLKjWlXLF6rzwSsYXD1sL+cYZKAV4bdaewzz3LuBSUKIRCGEG3A9sK7bmnXAbebXVwNfSDnYBjKD40BuFcnR/hiSzlEbyk+N5Ok0o53qAnj9WlUIdvfnMHWV2j51FUTNUzU/S34Alz8PwNmTQ7lwRji/uXym1aJRD1cjL94yn8gAdQE7c5K55sDNi+bACUwXWRwt6Gr9bj9VhknCIxdOGR8KA1QcAtQY1sESs0DN9S48rN5X54NflO1kG+MMdDpL5zQCD+B8YB/w6lBPLKVsFULcD3wKGIGXpZSpQohfAXuklOtQlej/EUKkoyyM63s/4vBpbjVxLL+aO5YmqPQ9D38o00pD0wstDfDWDcqFeddnEDatY58QcMv70NoMvuHtm40GwT9uSenzsME+7rx+12L2ZJUTFdDRgt81ejbTy7/klfxqrpjbsX77qTL8PFyYFukHDZUqdXysU5SqhiP5DsEVF2NOe87ZpQLj1XnaNTUIBqQ0pJQ/6PxeCOEP/Ge4J5dSbgA2dNv2eKfXjajhT3bheGE1zW0mZscGqB990ARtaWh6Z8/LypK4/g2IsDKD23Poqa9xwV49CgQNkbOIPvIuxzKygOnt23ecLmNhYjDGvS/B+h/Dze/BxPOHfO5RQVEqhM8YWoaTX6Qq8rNYK9X5Or12EAy1y2090NMhO8qxzHueFeOvNgQlqVQ8jaY7JhPs+ifELu5wSY00ETMBaCs43N4uJL+ygayyeu522QDrfwTIDrfLWMVkguKjKtlgqIRNU8doaVT9wLSlMWAGGtP4SAixzvz4GDhBz/jDqOdAThUhPm5EW1wCwROgKrcjYKbRWDi1WQViF95jv3OaM6hmGDJZu19l/OzIKCOSMhan/Ul1z/UKHvvWccVpaKlXlsZQCZ2mZptU5ar3WmkMmIHGNDq3DGkFsqSUuSMgj0M5lFvJ7JiAjgFLQRNAmtSc4dApDpVN42TselG5OKZdZr9zeoeAbxTL2gp47EAeDy6fxPZTZZznYR5nuuxnsP5hKD9tP5kcQVGqeh6O0gibphRPtjk7UgfCB8xA3VMXSym3mh/bpJS5QognR1QyO1PT2EJ6SS2zOleAB09Qz52D4VIq81gzfilNV8OS5t+h8v7tSfwSFrTsJb+smhe/ymDLiRIu9klTQeGw6epGZ6wnbxQfBQSETh36McLMMaH0z9WztjQGzECVxgor21Za2TZqaTNJHjh/EudO7VTnEWQuRels7n/xa/j7UvsKp7EfhYeh+Fjfa7Y8Aa6eqh7D3sy6DveWKpa7HuJ3G48jpWS+TIWEpapeISgJavKhuf/ZG6OWoiPqhs5tGN0aLJ4Dyxxwv8jhyzVO6NM9JYT4LvA9IEkIcajTLl9g8MOLnZgALzceXD6560avIPAI6Lhza6qBnS+q2c3NdeA2Tto1jAdam1SF8bd/VXedDxyyXjSWvx9S34ezHwGfMPvLOeFc8ArhVyFHuPqMezk7rAHXv+VCwg/V/qBE9VyRCeHTez3MqKYotT0pYMh4+KnZJlU56jeuf8sDpj9L4w3gUlTQ+9JOj/lSyptHWDbnIHhCRwbVgTeVwgCozOn9M5rRRflp1eV0218gOkVdSLJ6GdSz+VfKFbTkB9b3jzRGV5h5NaF5Wzg/wR3XbPO9W8JZ6tniUh1rwXBTG7x5Azx/hvo9DidzyoKlrsY/ZvjHGkf0qTSklFVSykwp5Q2odh7nSSmzAIMQItEuEjqaILPSMJlU8NPdnI5bme1YuTTDI30zPL8E1lwGL56j7syvfxNuWwfufuoGoTtpm9REt7N+pAo/HcWsa1VF864XlXvFK7jDvx9o/lmOtVTx3D1wYoP6u0+7DGZcOfxjWv5mOgg+KAaacvsL4KfAz8yb3IDXRkoop8KSdvvh91QvqrN/rLZXZjlWLs3QMbXBJ4+qOQwtDapC+L6vYOrFKlYx43I4+iE01XZ8pqURNjyihnQtvNdxsoNqTxKzUMVWDr+jmh1aXGmeAUqJjLVg+In1qlXLDW/Bdf+BkInDP6YlGK6VxqAYaCD8CuAyoA5ASpmPimuMfeIWqy/r8Q2qz83Ce1VrZa00Ri+H/quaC178FNy9CW5+FwITOvbPvhFa6mDfq6oNSFsrfPO0qg+4+I/2z5jqjhBwxwZlGc2+ERZ9p+v+oAmjw9IoOAQbf9oxd6Qvjq9XytGWLVIs7imdOTUoBlqn0SyllEIICSCEGD9Ro6Rl8L8lXdsV+Mdq99Roo65U1TCETlWup8g5MO1S62vjFqvir09/BpseB9mm6nWmr3aeEaBGV2UZTb24576gJMjsJSbjLNSXw1s3qvhRyl0QOrn3tSVqSFUP5ThcwqapWSZjveWKjRmo0nhbCPEPIEAIcQ9qvsa/Rk4sJ6N7f5uAOKjQlobTUVem3EvWUjEPvKFcTuIjpQAu/UvvfYssd/IZX6oW2kZXCIiHZBv40e1BUBIceku53lw9VaA/8xvnGS5kMsHa+5TCAKUQ+lIaxz9Wz1NsnOXv4q5mmWgGxUAbFj4lhFgBVANTgMellJtGVDJnJjBepV5qnAeTCV5aDgi4fb3Ku68rU2nTQqgpb1Fz4eb31cUqcnbfx/MKUkpitCiKzlgyqE5/DZGzVLC/KltddL1DHCsbQNqnaljVsv+BL3+rlEZfnNioLEOd5eQUDNTSwKwkNgEIIYxCiJuklK+PmGTOTEAcNJSrug338RHacXpydpr9+ALWXKLuttM+gzMfgrm3QMEBuOA3Shl4DXdSsZOTtEx9R9+8Xvnrq8yu1PLTzqE0cnerOOHSB1QGWFla72tNJmXtLXJw8oGmnT4D4UIIPyHEz4QQzwohLhCK+4EM4Fr7iOiEBMSpZx3XGBma6+DjhwdXC3PkXXDxVEHtmkJlCcYsgG1/VVX8ADOuGBl5nQ3vEPjONyoLrLYQLvyt2u4swfH8Aypm5OoBIZP6zvSqLVTpxYHjI8N/NNCfpfEfoALYDtwNPIJKt10tpTwwwrI5LwHx6rkye3hN0zTWOfI+7HlJVWlf/hxU5cHBN+CMH1gfydnWCqkfwJSLYOJyePCwqvBtqYdnF0LqWohdNL7cGx7+cPXLKq6BgE8fU9lfjkZKZTlMvki9D56g6l96oyJTPXfObtM4lP5SbpOklLdLKf8B3ACkAJeMa4UBXZWGxvYceEM9H/qvUhgffBe++A1882fr609vVTUXyVep915BKsjpGahSZAFm2m2Wl3Ph6qkUrV+0c1ga1fnq/ypqjnofPBFqi6Cx2vp6rTScjv6URovlhZSyDTgtpawZWZFGAd4hyhXSOYOqoQL2/QfevhXSPnecbKOdslOQ/a3qICtN8Ma1Sin4x6laidJOQdMdL8Bzi+DD+1UV90QrfTVnXA53b1bHG88EJTpHy/QC8/2mJREh2Fyk11swvCIThEGluWucgv6UxmwhRLX5UQPMsrwWQvRyazAOEELFNToX+P17Fay7H459BJ/9XJnhmsFz8E11kTjnJypzqeiIik3c9ZkqqtzwI/W3bWmAL38PbS3KL37u/1h3XYGaA20ccM7H2MRZplAWHFT/v5beUe1Ko5e4RkUm+MU4vqBS006fvyQppdFegow6AuM7LI2WBtXj/4z7VcHQh99XOf7OUgg2WjCZVOFd0rmqtcM5P4XqAlj1lEqhPe/nsPEnqgdRUw00VqqWEolnO1py5ycoUbmFGqtVh1dHkX8AQqZ01NIEJQGiq6VxaotKhph2iVIaQQkOEFTTG0OdEa4JmaxSBU1t5i+8hOh5kHw1eIcq14lmcOTtgepcmHWdeh8yCe5Y39HuIeUudcH57OcqVTN4Ykd3V03fWGbD2DsYfvpraKjseF9wsCOeASr2FBDXVWl88Wv4+CFlUVZk6niGk6GVxlAJnwGtjcrkLzGP2wyZolwkKXepAqbSPvLPNT05vh6EESZfYH2/0QUufEL9zfP2qjhFb1Xdmq60d7+1o9LI2KpqZraah3zWFKoU2u6FlcETO2o1TCYoPg51xZC3TwXJtdJwKrTSGCqWVNuiI6r5nTB0+GcX3AWu3vD2baoqWTMwTmxQE+g8A3tfM3E5TDhPxTfm3Gg/2UY7luFM9oprtDTAxw+q1yH0ALEAACAASURBVKkfKGVw8lP1PmZh17XBE1VMw2RS1fotdWr7vlfUs1YaToVDlIYQIkgIsUkIkWZ+7nGVEELMEUJsF0KkCiEOCSGuc4SsvRIyRd0VFx1VlkZAfEcg1icMbnhDDcL5z+re0wk1HZSmK+U7ZVXf64SAq16CuzaN/cpuW+Luq9ym9nJPffVHpaDm3qLGz+buhn1rVMPI6Hld10bOguZa9f9vGbVrcIHD76nXWmk4FY6yNB4FNkspJwGbze+7Uw/cKqWcAVwE/EUIYcO+yMPE1UPdIRWlqi+7ZeawhaRlcM0rauZ06loHCGhnsnfC3+b3nm584A1Y18e0uxPr1bO1rq3d8QpSFxrN4AhKso97qq0Vdvxd1c1c+FswuqvZH3l7Yd5tPV2KFssjdxeUmJXG9NUdFoeuBncqHKU0VgNrzK/XAJd3XyClPCmlTDO/zgeKgVC7STgQwqdD4SFzl84pPfdPuhDcfJQLayxTmq76HJWlw/v3qKFVnSnPUG1B9r3acSfZneMbIHxmR4sWje0JtFOtRulJdcGfdKHK1Jq0QtXaGN1g9vU91wdPVC7JnF3q++EX09G23t2vb3elxu44SmmESykLAMzPYX0tFkIsRLUvsZrMLYS4VwixRwixp6SkxObC9kr4DOWDbWtW7qruGAxqOlhRqv1ksjeNVfD6VSqmc9N76m/x7l2qfgJUBsy6Hyp3gzDC4Xd7HmPXPyFnB8y8yr6yjzf8IlUgeqRriCwFfJYsqenme8Jpl1l3KRoMqhYnd7dKXQ+bCglnA0KltutkB6dixJSGEOJzIcQRK4/VgzxOJKoH1h1SSpO1NVLKF6WUKVLKlNBQOxojYZ36TlmzNAAikqHwyNgt9tv0uGqncv3rMGk5XPqMUgBbnlD7tz8LmV/DBb9WLrvD73T9Wxz7SI1RnbxS9ZbSjByeQWBqVfGDkST/gEoEsSSGTL1YWQ5nPtT7Z2IWQslxFR8MmwbewWpSX9TckZVVM2hGrExWSrm8t31CiCIhRKSUssCsFIp7WecHrAd+LqXcMUKiDp3OzQpDJvW+Zs/LymUTMMZaIWRshb2vwJIfqGl3ADOvVq6Ib/6sEgD2vKQuGPNvVxlPH3xH3VHGLIAdz8Nn/wvR81VzvfFetT3SWNw8DRUj29K/4ICKORnMtcFu3nDda31/JtYc12hr7pjdfctaZcFqnApH/Y+sA24zv74N+LD7AiGEG7AWeFVK+Y4dZRs4AXHg5gu+kaqrqDUs7RLGmovKZIKPHlDB1WX/03XfRU+qH/6el1R191UvKRfD1FVKcXzyM3jpAvj0f9RgoFvWWp+2p7EtFqVRXz5y52hrVckfkXP6X9uZ6PkdCiJ0qno2unYoHo3T4Cil8XtghRAiDVhhfo8QIkUIYRkjey1wNnC7EOKA+THIb+III4S6Q4qe3/say11TX8HwtlZorretbCNN6UmVvnnmQz0v+G5eyl119iPq2cVdbffwUxk1BQfUHeWFv1N3oI5sazGesMQTGipsf+y9a2D78+YgeH3Xqu+B4O5jdveK3l29GqfAIf4AKWUZ0GOau5RyD2puB1LK14B+bFon4NpX+w7UefipGo7elEbGVlUE1dYK936pfLmg2pO8sBTm3qTcP85G9nb1HL/U+v6gJNUrqjurn4NL/qIb0DmCzu4pW9LWAp//Qh133q1q22AtDVCWqIefcmdpnBbtMBwu7j79f8nDk627pw68Ca9eplqA1xaqdFWTOdafvV3lrKdvtr3MtiB7O3iHdfQ0GihCaIXhKEZKaWRsVcd08VBp1a7evcf4+uLcn8EdG2wrm8bmaKVhDyKSVQ1DS0PX7Sc/UXMCvrcDLvo9nNqsZkaAar0Aqg7EGTOvsrer4LdOhxw9tCsNG8c0UteqeoprX1XvI2bqWMQYRisNexA+Q1kT3QvbStPUPldPSLlT5bNv/YPqw3NsnSqGqi+D6jzHyN0bVXkqzTbuDEdLohkMLu7KCujcdXa4tDbD8Y9gysUw+UJY+Ye+U2s1ox6tNOyBpfCvc/tnU5vqTWUx44VQLRcMRlVdXVukuuUCFByyr7z9kWPOfrak2WpGD56BtnVPZWxRBZ7JV6r3i+5Ts9o1YxatNOxBYAJq0EyngvaqHNVaPbiT79c/GpY+qDJQXDzgrIfV5wqdTGlk71B3rBG6/9OowyvQtim3qWtVunmSHjg2XtBKwx64eqjYRXknpWGZtREyuevaJT9Q2VZTL1HdcoMnOpelUVMEaZ9B7AJdjDcasaWl0fL/7Z15eFRVtuh/KwMkQsIQkMEIQRSCJKECgRsugwgijRdosW1bGwU+7FYez6f2/aAFbRF57WtU7nv3adPa6kUQeYg2Cor0FURpBr1gmGcBiZCAkAQSxgAJ+/2xTyVFqEoqQ1VlWL/vy3fO2WefXevs1Kl19tprr1Voc6AkjlTnhgaEPvXBIu6Wa0cavpRGoxtg4no7nwF2Ze2RjcGRsSIOfQ1LfmPDUPxsVqilUapCdAub5KgmOPQVXDoD3UfXTHtKnUBHGsHCnWjG7QmV+719gN3rMjyJii3NzdGuh02BGshVvP7y+b9CdHN47B92JbdS94huWXPeU7s/sd/hW+6omfaUOoEqjWDRsjNcKrDeUGBHGmVHGd5wzxsc3x442fzhfK4Ncd5zrI1CqtRN3Oap6rpxX7loMy12G2nDfSgNBlUawSKus926TVR5B/xbANWuh43Js+7fQps6NivDbuN7h04GpfpEt6iZSLcHv7RtqGmqwaFKI1i0dJTGqUPWT/7ciWs9p3xxQ0sY9bpNUPPXAXZ9RCjI+s7mw6hKeAil9uCOP1Vdc+cPa+yCvoSB1RZJqVuo0ggWLTraH928g6XrNfwxTwGkPmzDK5zJtvknQkHWd3Zlu0ajrdvUVCiRnP02sKB60DU4VGkEi/BIqzjyDtlkM+C/0gCIT4Mb4kqvDSZXiyF7i5qm6gM1rTSUBoe+JgSTlp1tOsufdtpcHC0SKnd960TI+T4gonkle7OVNb43XD6rSqM+EO0Oj14N89SFU3D+pPcUx0q9R5VGMInrDAdX2f1xn1V+aN+qi3VzNCbwgQILC2DRr2303dbdbJkqjbpPTYw0cp0XFx1pNEjUPBVM3JPhfR6DTlWYQGzdFQrz4XxOzcrljS9n2LfJ7vfZEO3RLSofBl2pfdSE0sjZb7eVMa8q9QYdaQSTbiPtQr07nqna9e43u5z9NsRIoPjxW5vXvO8TcPcf4eZ/soEUNQx63SeiETRqCheqOdKIiLYmVqXBoUojmMS2g6Ezq36924acsw86DagZmcpy6SwsnWjjX935rFUU6RMD81lKaKhu/KmcfdDqVs2Z0UBR81RdIrY9NIoptSkHgi+etWtBRv9V027WV6KbV1NpfK+T4A0YVRp1CRFo3SVwbrfbF9t0nf2eho6aYKneUp34U5fPQ8ERnQRvwKjSqGsEyu1259+sWSphAAyaVvPtK7WHG+Kq7kzhjs6sSqPBokqjrtGqi3WDrcmUnce2wse/hQ7/DL9erLkR6jtxneH0j1B0yf9riotsNIJvXrPHap5qsIREaYhISxFZJSIHnG2LcurGiki2iPw5mDLWWtxveDU5r/H9Srv241cLdB6jIdA6EUzxtfldKmL3x7D4YdizzI5G3QE4lQZHqEYaU4HVxpjbgNXOsS/+J/CPoEhVF/B0u60pjm6EG28vDWan1G/c6ysqMzd25FsboHBaNoxfruHQGzChUho/B+Y7+/OBe71VEpFeQBtgZZDkqv007wjhjWtuMvxqsQ1G2OGfaqY9pfbT6jZAKjdaPfod3NSrNDmY0mAJldJoY4w5DuBsr1upJiJhwL8BU4IsW+0mLNy+KdaUeSpnn03ZebMqjQZDZLSNe+bvi8elc3ByN9zcJ6BiKXWDgC3uE5EvgbZeTj3nZxOTgBXGmKNSwUpkEXkMeAygQ4cGsEq1dRc7OqgJjvyX3arSaFi07uq/ifPYVjBXNfaYAgRQaRhj7vJ1TkROiEg7Y8xxEWkHnPRSrS8wQEQmAU2BRiJyzhhz3fyHMeYt4C2AtLS0auaxrAO0ToRdH8PlC77zWxz80nrHJP5L+W0d3QhN21Q+4q5St2ndFQ6utl5RFQXOzNpktzf1CrxcSq0nVOapT4Fxzv44YFnZCsaYMcaYDsaYBGAy8J43hdEgadUFMDZlrC9WvQD/6cd6i6MbrdlB40o1LFonwtUrcDqz4rpZGRB3qzpKKEDolMYsYKiIHACGOseISJqIvBMimeoOFXlQXb4AJ/dC/o/lp/U8e8L+aNycXuMiKrUczzhmmet9u98aY02h8TqfoVhCErDQGJMHDPFSngH8xkv5PGBewAWrK7TsbFPH+lIaP+20fvgAx7dB58He653YabftU2teRqV209pxu/3mNZt/vttIu06nLKcz7erx+LSgiqfUXnRFeF0kopHNbeHL++XYFo/9bb7byfvBblvdVnOyKXWDxjEQG2/Nk5jSvPVlObnXbtv1CJpoSu1GQ6PXVVp39e12m70FYtpBRGM70vBF3kEbNbdJ68DIqNQqrly5QlZWFoWFhbZg0Nt2nU5YmA1EuGfP9XNbl1rBsA+hIArO7g2+0EqNExUVRXx8PJGRVVugqUqjrtK6K+z/O8wdbifEn9pR6kl1bAu072lX7XqOOspy6pANB6GT4A2CrKwsYmJiSEhI4Bo39vO5UHAUbrz1+rhjBVlwoTG07a7fk3qAMYa8vDyysrLo1KlTldpQ81RdpW2ynbc4tsXanPOP2PKL+XYEcVOqnavIP+J7MjzvoMYQakAUFhYSFxfHdeueIhrbbbGXAIZFlyG8kSqMeoKIEBcXVzrarAKqNOoq3UbB+BXw6w/t8Zlsu3Wbo9r3hPaua8s8KbpsFUrcrYGXVak1eF0oG+4ojSIvPyTFjtJQ6g0VLZauCFUadZWwcEjoBy062uOzx+022zFHtU8tnbw8tvX6609n2lW+LXWk0eAJjwTEvkh4YoxVGjUYKl9EeOSRR0qOi4qKaN26NSNGjABg3rx5PPHEE+W2MWjQILp27UqPHj3o168f+/dXLnjnvHnzOHbsWMnxunXr6N69Oy6Xi4sXL3q9JjMzk6SkJAAyMjJ48skn/f68NWvW0KxZM1JTU0lMTGTy5MkVXrNt2zZWrFjh92cEE1UadZ2YdnZ7xnkITuyC5h3sQqzoFtCiE/z4zfXXnXL88nWkoYhYE1XZkcbVYmsCdY9EaoAmTZqwa9eukh/nVatWcdNNN1W6nYULF7J9+3bGjRvHlCn+h6crLi6+TmksXLiQyZMns23bNqKjoytsIy0tjddee61S8g4YMICtW7eydetWli9fzoYNG8qtXxWlYYzh6tWrlbqmKqjSqOtENLbeT27z1Kkfrh09JN9vw0Wc+uHa69wulnG3BEdOpXYT0fj6OY1iZ+RRw+ap4cOH8/nnnwOwaNEiHnrooSq3NXDgQA4etN/l1atXk5qaSnJyMhMmTODSJXs/CQkJzJw5k/79+7No0SIyMjIYM2YMLpeL119/nQ8//JCZM2cyZswYjDFMmTKFpKQkkpOTWbx48XWfuWbNmpKR0alTp7j33ntJSUkhPT2dHTt2lCtvdHQ0LpeL7Gz7vJ4/f54JEybQu3dvUlNTWbZsGZcvX2b69OksXrwYl8vF4sWLmTFjBrNnzy5pJykpiczMTDIzM+nWrRuTJk2iZ8+eHD16lKZNm/Lcc8/Ro0cP0tPTOXHiRJX71xvqPVUfiG1fOtI4dRiSflF6Lu1RWP9/YONbMHxWaXneIZv2M9pn/iulHvPiZ7vZc+xMaUHxJSi+Ao0KSsuuFtnRR+QFkIrfL29vH8sLI7tXWO/BBx9k5syZjBgxgh07djBhwgTWrVtXldvgs88+Izk5mcLCQsaPH8/q1avp0qULY8eO5Y033uDpp58GrJvp+vXrAXjnnXeYPXs2aWl2weLmzZsZMWIE999/P0uWLGHbtm1s376d3NxcevfuzcCBA31+/gsvvEBqaipLly7lq6++YuzYsWzb5tvN/fTp0xw4cKCkzZdeeonBgwczd+5c8vPz6dOnD3fddRczZ84kIyODP//Z5p6bMWOGzzb379/Pu+++y1/+8hfAKqL09HReeuklfv/73/P222/zhz/8wf9OrQAdadQHYm+ySuPCKSjMh5YernSx7aD7fbD1fSj0+JHIO6jzGUopbqVgPMwb7v0a9pxKSUkhMzOTRYsWcc8991SpDfdIYcOGDcyePZv9+/fTqVMnunSxK93HjRvH2rVrS+r/6le/8qvd9evX89BDDxEeHk6bNm244447+O473xGl169fXzJHM3jwYPLy8igoKLiu3rp160hJSaFt27aMGDGCtm1tAPCVK1cya9YsXC4XgwYNorCwkCNHjvjdDwAdO3YkPb00FFCjRo1KRkK9evUiMzOzUu1VhI406gOx7W1mtdOH7XGLMv7X6RNh54ew+V3o95QtO/UDdLojuHIqtYbrRgSXztn1Pi07Q1SsLcs/ChdPQ7uUGv/8UaNGMXnyZNasWUNeXl6lr1+4cGHJSAGosI0mTfxLY2xM5YJke6vvzTtpwIABLF++nO+//57+/fszevRoXC4XxhiWLFlC167X5lzfuHHjNccRERHXzFd4usyWvbfIyMgSGcLDwykqKqrUPVWEjjTqAzHt7MPtDvnQsozSuKkX3DoUvv5fcGI3nP3JzoHofIbixr1Wo8hjXiOA7rYTJkxg+vTpJCcn10h7iYmJZGZmlsxvLFiwgDvu8P5SFBMTw9mzZ72eGzhwIIsXL6a4uJicnBzWrl1Lnz6+gzUOHDiQhQsXAnauo1WrVsTGxvqs36VLF6ZNm8bLL78MwLBhw3j99ddLlM/WrVu9ypiQkMCWLdYzcsuWLRw+fNjnZwQaVRr1gVjH+8TtJeUtN8a9f7E5nhc/Am/dCRFRcKvPlCdKQyMswpqoissojRp0t/UkPj6ep556yuu5efPmER8fX/KXlZVVYXtRUVG8++67/PKXvyQ5OZmwsDAmTpzote748eOZOHGiVxfb0aNHk5KSQo8ePRg8eDCvvPJKiSnJGzNmzCAjI4OUlBSmTp3K/PnzfdZ1M3HiRNauXcvhw4d5/vnnuXLlCikpKSQlJfH8888DcOedd7Jnz56SifBf/OIXnDp1CpfLxRtvvFFihgsFUtnhWG0nLS3NZGRkhFqM4PLDP+C9UTZ/eFEhTPYRk+qHNfDevXZtxwMLAmJ2UGove/fupVu3br4rnNxrRxwtb7FrNH7aYZ0lmsUHT0glKHj7LojIZmNMheGMdU6jPuAeaeT/CB36+q53yyCYuM6u44hqFgzJlLpEeKT1oALrOWWu6mpw5TpUadQHYtuV7pedBC9L25qxISv1kPBGcMXx/HGv0YiouYV9Sv1A5zTqA42aQFRzu192ElxR/CU8snSE4Z4Qr8HV4Er9QJVGfSG2vd1WNNJQFF+4TVHFV0onxNU8pZRBlUZ9wa00dKShVJUSpXHZBi8Mi7QJmhTFA/1G1Bd0pKFUl3Ank1vxZWue0vkMxQuqNOoLt9wJnYfY6LaKUhXCypinAqA0wsPDcblc9OjRg549e/LNN14iMFcDz2CCgaZp06YV1nHfb1JSEiNHjiQ/P7/c+vn5+SUxpGorqjTqC0n3wSMfa4Y1peqEhdlFfkWFdkI8APMZ0dHRJQEB//SnPzFt2rQa/4xAU5kQ5O773bVrFy1btmTOnDnl1q+q0iguLq70NVVFlYaiKKWER9o4VBBw89SZM2do0aI0yvKrr75K7969SUlJ4YUXXgAoCf3929/+lu7du3P33XeXrOI+ePAgd911V8mo5dAhmyPm3Llz3H///SQmJpaEOwcbiuPZZ5+lb9++pKWlsWXLFoYNG0bnzp158803S64dMmQIPXv2JDk5mWXLll0jh2cIcje5ubn07du3JNy7L/r27VsSEt3X/U6dOpVDhw7hcrmYMmXKdSOnJ554gnnz5pXcjzvk+0cffcSgQYN45pln6NOnD126dKly5OCKCMk6DRFpCSwGEoBM4AFjzGkv9ToA7wA3Awa4xxiTGTRBFaW+8vep8NPO68uLLtpRBkDkDSDh/rfZNvna8PteuHjxIi6Xi8LCQo4fP85XX30F2GivBw4cYNOmTRhjGDVqFGvXrqVDhw4cOHCARYsW8fbbb/PAAw+wZMkSHn74YcaMGcPUqVMZPXo0hYWFXL16laNHj7J161Z2795N+/bt6devHxs2bKB///4A3HzzzXz77bf87ne/Y/z48WzYsIHCwkK6d+/OxIkTiYqK4pNPPiE2Npbc3FzS09MZNWoUcH0IcoATJ04watQo/vjHPzJ06FCf911cXMzq1at59NFHy73fWbNmsWvXrpLw6mvWrCm3Pz1Dvr/55psUFRWxadMmVqxYwYsvvsiXX35Z7vVVIVSL+6YCq40xs0RkqnP8jJd67wEvGWNWiUhTIPBpqRSlIeOZN8OPHBqVxW2uAfj2228ZO3Ysu3btYuXKlaxcuZLU1FTAvvEfOHCADh060KlTJ1wum+/eHer77NmzZGdnM3r0aMD+eLrp06cP8fE29InL5SIzM7NEabgVQHJyMufOnSMmJoaYmBiioqLIz8+nSZMmPPvss6xdu5awsDCys7NLkhiVDUF+5coVhgwZwpw5c3wGR3QryczMTHr16lWiWMq738pQNuT7fffdd00/BYJQKY2fA4Oc/fnAGsooDRG5HYgwxqwCMMacC6J8ilK/8TUiOHfC5maR8IDHJuvbty+5ubnk5ORgjGHatGk8/vjj19TJzMykceNSM1l4eDgXL14sN4R52fqeocHd58LCwq6pFxYWRlFREQsXLiQnJ4fNmzcTGRlJQkJCSRjysiHIIyIi6NWrF1988YVPpeFWkgUFBYwYMYI5c+bw5JNPlnu/ZT/DV0h0bzK57ykQIdHdhGpOo40x5jiAs73RS50uQL6IfCwiW0XkVZHKjJUVRak07snvILjb7tu3j+LiYuLi4hg2bBhz587l3Dn7bpidnc3Jkyd9XhsbG0t8fDxLly4F4NKlS1y4cKHaMhUUFHDjjTcSGRnJ119/zY8//uizrogwd+5c9u3bx6xZ5ZvlmjVrxmuvvcbs2bO5cuWKz/stGxK9Y8eO7Nmzh0uXLlFQUMDq1aurfY/VJWAjDRH5EvAWU/g5P5uIAAYAqcAR7BzIeOA/vHzWY8BjQKWHd4qieOBWGgEKH+I214D1Qpo/fz7h4eHcfffd7N27l759bcDNpk2b8v777xMe7vs9ccGCBTz++ONMnz6dyMhIPvroo2rLN2bMGEaOHElaWhoul4vExMRy64eHh/PBBx8wcuRIYmNjmTRpks+6qamp9OjRgw8++IBHHnnE6/127tyZfv36kZSUxPDhw3n11Vd54IEHSElJ4bbbbisxZ4WSkIRGF5H9wCBjzHERaQesMcZ0LVMnHZhljBnkHD8CpBtj/nt5bTfI0OiK4gcVhkYHuxL85G5o2qZ0wahS76hOaPRQmac+BcY5++OAZV7qfAe0EJHWzvFgYE8QZFOUhkt4JMS0hWhdJKp4J1RKYxYwVEQOAEOdY0QkTUTeATDGFAOTgdUishMQ4O0QyasoDQMRmz44MqriukqDJCTeU8aYPGCIl/IM4Dcex6sATS+nKIpSS9AV4YrSgKhv6Z2VylPd74AqDUVpIERFRZGXl6eKowFjjCEvL++axZCVRdO9KkoDIT4+nqysLHJyckItihJCoqKiSlbMVwVVGorSQIiMjKRTJ823olQPNU8piqIofqNKQ1EURfEbVRqKoiiK34QkjEggEZEcwHeUsYppBeTWkDg1icpVOVSuyqFyVY76KFdHY0zriirVO6VRXUQkw5/4K8FG5aocKlflULkqR0OWS81TiqIoit+o0lAURVH8RpXG9bwVagF8oHJVDpWrcqhclaPByqVzGoqiKIrf6EhDURRF8RtVGg4i8jMR2S8iB0VkagjluFlEvhaRvSKyW0SecspbisgqETngbFuESL5wJ2f7cue4k4hsdORaLCKNQiBTcxH5m4jsc/qtb23oLxH5nfM/3CUii0QkKlT9JSJzReSkiOzyKPPaR2J5zXkWdohIzyDL9arzv9whIp+ISHOPc9McufaLyLBgyuVxbrKIGBFp5RyHtL+c8v/h9MluEXnFo7zm+8sY0+D/gHDgEHAL0AjYDtweIlnaAT2d/Rjge+B24BVgqlM+FXg5RPL9K/D/gOXO8YfAg87+m8B/C4FM84HfOPuNgOah7i/gJuAwEO3RT+ND1V/AQKAnsMujzGsfAfcAf8cmPksHNgZZrruBCGf/ZQ+5bneezcZAJ+eZDQ+WXE75zcAX2LVgrWpJf90JfAk0do5vDGR/BfzLWhf+gL7AFx7H04BpoZbLkWUZNrvhfqCdU9YO2B8CWeKB1djUu8udhyTX4wG/ph+DJFOs8+MsZcpD2l+O0jgKtMQGBl0ODAtlfwEJZX5svPYR8FfgIW/1giFXmXOjgYXO/jXPpfPj3TeYcgF/A3oAmR5KI6T9hX0RuctLvYD0l5qnLO4H3E2WUxZSRCQBSAU2Am2MMccBnO2NIRDp34HfA1ed4zgg3xhT5ByHot9uAXKAdx2z2Tsi0oQQ95cxJhuYDRwBjgMFwGZC31+e+Oqj2vQ8TMC+xUOI5RKRUUC2MWZ7mVOh7q8uwADH7PkPEekdSLlUaVjES1lI3cpEpCmwBHjaGHMmlLI48owAThpjNnsWe6ka7H6LwA7X3zDGpALnsaaWkOLMD/wcaxZoDzQBhnupWhvdF2vD/xUReQ4oAha6i7xUC4pcInID8Bww3dtpL2XB7K8IoAXWNDYF+FBEJFByqdKwZGFtlW7igWMhkgURicQqjIXGmI+d4hMi0s453w44GWSx+gGjRCQT+ABrovp3zUSb4QAAA69JREFUoLmIuPOyhKLfsoAsY8xG5/hvWCUS6v66CzhsjMkxxlwBPgb+mdD3lye++ijkz4OIjANGAGOMY1sJsVydsS8A251nIB7YIiJtQywXzud/bCybsJaAVoGSS5WG5TvgNsezpRHwIPBpKARx3hD+A9hrjPnfHqc+BcY5++Owcx1BwxgzzRgTb4xJwPbPV8aYMcDXwP0hlOsn4KiIdHWKhgB7CHF/Yc1S6SJyg/M/dcsV0v4qg68++hQY63gFpQMFbjNWMBCRnwHPAKOMMRfKyPugiDQWkU7AbcCmYMhkjNlpjLnRGJPgPANZWIeVnwhxfwFLsS9xiEgXrDNILoHqr0BN1tS1P6wHxPdYD4PnQihHf+wQcgewzfm7Bzt/sBo44GxbhlDGQZR6T93ifBEPAh/heHAEWR4XkOH02VLsUD3k/QW8COwDdgELsF4sIekvYBF2buUK9gfvUV99hDVrzHGehZ1AWpDlOoi1xbu//2961H/OkWs/MDyYcpU5n0npRHio+6sR8L7zPdsCDA5kf+mKcEVRFMVv1DylKIqi+I0qDUVRFMVvVGkoiqIofqNKQ1EURfEbVRqKoiiK36jSUBQ/cSKbLvA4jhCRHHEi/lahveYiMsnjeFBV21KUYKFKQ1H85zyQJCLRzvFQILsa7TUHJlVYS1FqEao0FKVy/B34F2f/IexiK6AkP8VSJ6fCf4lIilM+w8mDsEZEfhCRJ51LZgGdRWSbiLzqlDWV0twgC53V5IjILBHZ47Q9Ozi3qijXE1FxFUVRPPgAmO6YkVKAucAA59yLwFZjzL0iMhh4D7taHSARm/cgBtgvIm9gAysmGWNcYM1T2KjG3bExgjYA/URkDzZEeKIxxngmJVKUYKMjDUWpBMaYHdh8Bg8BK8qc7o8NF4Ix5isgTkSaOec+N8ZcMsbkYgMDtvHxEZuMMVnGmKvYEBoJwBmgEHhHRO4DLvi4VlECjioNRak8n2JzZSwqU15eKOpLHmXF+B7lX1fP2PwbfbCRj+8F/rOyAitKTaFKQ1Eqz1xgpjFmZ5nytcAYKDE15Zryc6GcxZqrysXJrdLMGLMCeJpSk5eiBB2d01CUSmKMyQL+r5dTM7AZBHdgTUjjvNTxbCdPRDaIyC7sBPvnPqrGAMtEJAo7mvldVWVXlOqiUW4VRVEUv1HzlKIoiuI3qjQURVEUv1GloSiKoviNKg1FURTFb1RpKIqiKH6jSkNRFEXxG1UaiqIoit+o0lAURVH85v8Dkns0fZskmcMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "temp = df['premium']\n",
    "premium = temp[int(len(temp)/2):]\n",
    "\n",
    "cumsum = 0\n",
    "cs = []\n",
    "\n",
    "cumsum_h = 0\n",
    "cs_h = []\n",
    "\n",
    "for i in range(len(predicted)):\n",
    "    if predicted[i] == 1:\n",
    "        cumsum += premium[i+884]\n",
    "    else:\n",
    "        cumsum += -1*premium[i+884]\n",
    "        # unblock the above two lines to impose short-selling constraint\n",
    "    cs.append(cumsum)\n",
    "\n",
    "for i in range(len(bull_test)):\n",
    "    cumsum_h += premium[i+884]\n",
    "    cs_h.append(cumsum_h)\n",
    "    \n",
    "plt.figure()\n",
    "plt.plot(cs, label='ML Portfolio Return')\n",
    "plt.plot(cs_h, label='Benchmark Return')\n",
    "plt.legend(loc='best')\n",
    "plt.ylabel('Return ($)')\n",
    "plt.xlabel('Months')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Regression (Equity Premium Prediction)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "ret = df.premium\n",
    "x = df.iloc[:,4:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will first start with the simplest linear regression model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda3/lib/python3.6/site-packages/sklearn/cross_validation.py:41: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.\n",
      "  \"This module will be removed in 0.20.\", DeprecationWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE of regression:  0.008048333770433132\n",
      "MAE of hist average:  0.00515315257338971\n",
      "Diff MAE:  -0.0028951811970434217\n",
      "RMSE of regression:  0.00010014956285362834\n",
      "RMSE of hist average:  0.039741222058134955\n",
      "Diff RMSE:  0.03964107249528133\n",
      "adjusted R2: 1.0019532594410034\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.007740374177464245\n",
      "MAE of hist average:  0.005587463837539477\n",
      "Diff MAE:  -0.002152910339924768\n",
      "RMSE of regression:  9.386462957548477e-05\n",
      "RMSE of hist average:  0.040980263177127774\n",
      "Diff RMSE:  0.040886398547552286\n",
      "adjusted R2: 1.0017049841884225\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.010514353766484411\n",
      "MAE of hist average:  0.005697696477008201\n",
      "Diff MAE:  -0.004816657289476211\n",
      "RMSE of regression:  0.00019343503779616178\n",
      "RMSE of hist average:  0.04605211021996715\n",
      "Diff RMSE:  0.04585867518217099\n",
      "adjusted R2: 1.0027699005155797\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.00904529467310396\n",
      "MAE of hist average:  0.005065633839807874\n",
      "Diff MAE:  -0.003979660833296086\n",
      "RMSE of regression:  0.00013902260129058125\n",
      "RMSE of hist average:  0.040025618571566925\n",
      "Diff RMSE:  0.03988659597027634\n",
      "adjusted R2: 1.00275195931517\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.007989431247970146\n",
      "MAE of hist average:  0.005840597916020638\n",
      "Diff MAE:  -0.002148833331949508\n",
      "RMSE of regression:  9.652124190081015e-05\n",
      "RMSE of hist average:  0.04481924193582679\n",
      "Diff RMSE:  0.04472272069392598\n",
      "adjusted R2: 1.0014766546532778\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.009214248024801301\n",
      "MAE of hist average:  0.005056398443729396\n",
      "Diff MAE:  -0.004157849581071905\n",
      "RMSE of regression:  0.00015174926043499016\n",
      "RMSE of hist average:  0.04100392460096191\n",
      "Diff RMSE:  0.04085217534052692\n",
      "adjusted R2: 1.002770328139347\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.009354609921984712\n",
      "MAE of hist average:  0.005066526185353951\n",
      "Diff MAE:  -0.004288083736630761\n",
      "RMSE of regression:  0.00013549348745406999\n",
      "RMSE of hist average:  0.04022061091431747\n",
      "Diff RMSE:  0.0400851174268634\n",
      "adjusted R2: 1.0026126695331663\n",
      "-------------------------------------------------------------------\n",
      "the main take way: Mean(Diff_MAE) is -0.0035 Mean(Diff_RMSE) is 0.0417\n"
     ]
    }
   ],
   "source": [
    "# Here are 67 years of data prior to 2013 \n",
    "# Spilt the data into 7 folds and cross-validate\n",
    "# In each sample we have around 10 years of data\n",
    "\n",
    "pd.options.mode.chained_assignment=None\n",
    "import math\n",
    "from sklearn import linear_model\n",
    "from sklearn.metrics import mean_absolute_error,mean_squared_error, r2_score\n",
    "from sklearn.cross_validation import KFold\n",
    "\n",
    "kf = KFold(x.shape[0],n_folds=7, shuffle=True)\n",
    "\n",
    "\n",
    "diff_MAE = []\n",
    "diff_RMSE = []\n",
    "\n",
    "for train_index, test_index in kf:\n",
    "    x_train, x_test = x.iloc[train_index], x.iloc[test_index]\n",
    "    ret_train, ret_test = ret.iloc[train_index], ret.iloc[test_index]\n",
    "\n",
    "    regr = linear_model.LinearRegression()\n",
    "    regr.fit(x_train, ret_train)\n",
    "\n",
    "    ret_pred = regr.predict(x_test)\n",
    "\n",
    "    \n",
    "    diff = df['his_avg'].iloc[test_index] - df['premium'].iloc[test_index]\n",
    "    mae_avg = diff.abs().sum() / len(train_index)\n",
    "    diff_MAE.append(mae_avg - mean_absolute_error(ret_test, ret_pred))\n",
    "    rmse_avg = math.sqrt(np.power(diff, 2).sum() / len(test_index))\n",
    "    diff_RMSE.append(rmse_avg - mean_squared_error(ret_test, ret_pred))\n",
    "\n",
    "    r_squared = r2_score(ret_test, ret_pred)\n",
    "    adjusted_r_squared = 1 - (1-r_squared)*(len(ret_test)-1)/(len(ret_test)-x_test.shape[1]-1)\n",
    "    \n",
    "    print(\"MAE of regression: \", mean_absolute_error(ret_test, ret_pred))\n",
    "    print(\"MAE of hist average: \", mae_avg)\n",
    "    print(\"Diff MAE: \", mae_avg - mean_absolute_error(ret_test, ret_pred))\n",
    "    print(\"RMSE of regression: \", mean_squared_error(ret_test, ret_pred))\n",
    "    print(\"RMSE of hist average: \", rmse_avg)\n",
    "    print(\"Diff RMSE: \", rmse_avg - mean_squared_error(ret_test, ret_pred))\n",
    "    print(\"adjusted R2:\", adjusted_r_squared)\n",
    "    print(\"-------------------------------------------------------------------\")\n",
    "    \n",
    "print( \"the main take way: Mean(Diff_MAE) is\", round(sum(diff_MAE)/len(diff_MAE),4),\n",
    "      \"Mean(Diff_RMSE) is\", round(sum(diff_RMSE)/len(diff_RMSE),4))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Note:\n",
    "Diff_MAE = MAE of Historical Average - MAE of current method  \n",
    "Diff_MAE is Negative for linear regression: Linear model does a WORSE job on prediction.   \n",
    "Diff_MAE is Positive for ML methods: ML methods beat the benchmark forecast."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Random Forrest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda3/lib/python3.6/site-packages/sklearn/ensemble/weight_boosting.py:29: DeprecationWarning: numpy.core.umath_tests is an internal NumPy module and should not be imported. It will be removed in a future NumPy release.\n",
      "  from numpy.core.umath_tests import inner1d\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE of regression_in:  0.0004684632173754462\n",
      "MAE of regression_out:  0.0008318186605938289\n",
      "MAE of hist average:  0.00515315257338971\n",
      "Diff MAE:  0.0043213339127958815\n",
      "RMSE of regression:  4.233193552916601e-06\n",
      "RMSE of hist average:  0.039741222058134955\n",
      "Diff RMSE:  0.03973698886458204\n",
      "adjusted R2: 1.0000825617709876\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression_in:  0.0005188928816910979\n",
      "MAE of regression_out:  0.0007126055132834755\n",
      "MAE of hist average:  0.005587463837539477\n",
      "Diff MAE:  0.0048748583242560015\n",
      "RMSE of regression:  2.412315378283186e-06\n",
      "RMSE of hist average:  0.040980263177127774\n",
      "Diff RMSE:  0.04097785086174949\n",
      "adjusted R2: 1.0000438179918896\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression_in:  0.0004272013225987866\n",
      "MAE of regression_out:  0.0019163840033902444\n",
      "MAE of hist average:  0.005697696477008201\n",
      "Diff MAE:  0.0037813124736179563\n",
      "RMSE of regression:  0.00010841332366252862\n",
      "RMSE of hist average:  0.04605211021996715\n",
      "Diff RMSE:  0.04594369689630462\n",
      "adjusted R2: 1.0015524287871\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression_in:  0.0005332166130091775\n",
      "MAE of regression_out:  0.000785211644486592\n",
      "MAE of hist average:  0.005065633839807874\n",
      "Diff MAE:  0.0042804221953212825\n",
      "RMSE of regression:  6.157694556254521e-06\n",
      "RMSE of hist average:  0.040025618571566925\n",
      "Diff RMSE:  0.04001946087701067\n",
      "adjusted R2: 1.0001218918703632\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression_in:  0.0006004109199692137\n",
      "MAE of regression_out:  0.0014302234803197284\n",
      "MAE of hist average:  0.005840597916020638\n",
      "Diff MAE:  0.004410374435700909\n",
      "RMSE of regression:  1.7846160888207415e-05\n",
      "RMSE of hist average:  0.04481924193582679\n",
      "Diff RMSE:  0.04480139577493858\n",
      "adjusted R2: 1.0002730240100495\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression_in:  0.0005778989992019438\n",
      "MAE of regression_out:  0.0014235491761705001\n",
      "MAE of hist average:  0.005056398443729396\n",
      "Diff MAE:  0.003632849267558896\n",
      "RMSE of regression:  4.781416561369742e-05\n",
      "RMSE of hist average:  0.04100392460096191\n",
      "Diff RMSE:  0.04095611043534821\n",
      "adjusted R2: 1.0008728934037592\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression_in:  0.0004637569783428734\n",
      "MAE of regression_out:  0.001022581350992022\n",
      "MAE of hist average:  0.005066526185353951\n",
      "Diff MAE:  0.004043944834361929\n",
      "RMSE of regression:  2.8422153459861924e-05\n",
      "RMSE of hist average:  0.04022061091431747\n",
      "Diff RMSE:  0.04019218876085761\n",
      "adjusted R2: 1.0005480536061686\n",
      "-------------------------------------------------------------------\n",
      "the main take way: Mean(Diff_MAE) is 0.0042 Mean(Diff_RMSE) is 0.0418 In-sample avg:  0.0005 Out-sample avg:  0.0012\n"
     ]
    }
   ],
   "source": [
    "pd.options.mode.chained_assignment=None\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "\n",
    "diff_MAE = []\n",
    "diff_RMSE = []\n",
    "in_sample = []\n",
    "out_sample = []\n",
    "\n",
    "for train_index, test_index in kf:\n",
    "    x_train, x_test = x.iloc[train_index], x.iloc[test_index]\n",
    "    ret_train, ret_test = ret.iloc[train_index], ret.iloc[test_index]\n",
    "    \n",
    "    regr = RandomForestRegressor(max_depth=25, random_state=0, n_estimators=25)\n",
    "    regr.fit(x_train, ret_train)\n",
    "    \n",
    "    ret_pred = regr.predict(x_test)\n",
    "    ret_fit = regr.predict(x_train)\n",
    "    \n",
    "    diff = df['his_avg'].iloc[test_index] - df['premium'].iloc[test_index]\n",
    "    mae_avg = diff.abs().sum() / len(train_index)\n",
    "    diff_MAE.append(mae_avg - mean_absolute_error(ret_test, ret_pred))\n",
    "    rmse_avg = math.sqrt(np.power(diff, 2).sum() / len(test_index))\n",
    "    diff_RMSE.append(rmse_avg - mean_squared_error(ret_test, ret_pred))\n",
    "\n",
    "    r_squared = r2_score(ret_test, ret_pred)\n",
    "    adjusted_r_squared = 1 - (1-r_squared)*(len(ret_test)-1)/(len(ret_test)-x_test.shape[1]-1)\n",
    "    \n",
    "    in_sample.append(mean_absolute_error(ret_train, ret_fit))\n",
    "    out_sample.append(mean_absolute_error(ret_test, ret_pred))\n",
    "    \n",
    "    print(\"MAE of regression_in: \", mean_absolute_error(ret_train, ret_fit))\n",
    "    print(\"MAE of regression_out: \", mean_absolute_error(ret_test, ret_pred))\n",
    "    print(\"MAE of hist average: \", mae_avg)\n",
    "    print(\"Diff MAE: \", mae_avg - mean_absolute_error(ret_test, ret_pred))\n",
    "    print(\"RMSE of regression: \", mean_squared_error(ret_test, ret_pred))\n",
    "    print(\"RMSE of hist average: \", rmse_avg)\n",
    "    print(\"Diff RMSE: \", rmse_avg - mean_squared_error(ret_test, ret_pred))\n",
    "    print(\"adjusted R2:\", adjusted_r_squared)\n",
    "    print(\"-------------------------------------------------------------------\")\n",
    "    \n",
    "print( \"the main take way: Mean(Diff_MAE) is\", round(sum(diff_MAE)/len(diff_MAE),4),\n",
    "      \"Mean(Diff_RMSE) is\", round(sum(diff_RMSE)/len(diff_RMSE),4), \n",
    "      \"In-sample avg: \", round(sum(in_sample)/len(in_sample),4),\n",
    "      \"Out-sample avg: \", round(sum(out_sample)/len(out_sample),4))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Boosted Trees"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE of regression:  0.0008857335446538282\n",
      "MAE of hist average:  0.032321670148204605\n",
      "Diff MAE:  0.031435936603550776\n",
      "RMSE of regression:  3.967178712756121e-06\n",
      "RMSE of hist average:  0.04225215916481874\n",
      "Diff RMSE:  0.042248191986105986\n",
      "adjusted R2: 1.000077373570628\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.0007596410530923491\n",
      "MAE of hist average:  0.03188735888405484\n",
      "Diff MAE:  0.031127717830962492\n",
      "RMSE of regression:  2.1807679260101003e-06\n",
      "RMSE of hist average:  0.042054146292291074\n",
      "Diff RMSE:  0.04205196552436506\n",
      "adjusted R2: 1.0000396120972221\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.001782416353861496\n",
      "MAE of hist average:  0.03177712624458612\n",
      "Diff MAE:  0.02999470989072462\n",
      "RMSE of regression:  8.339333922440027e-05\n",
      "RMSE of hist average:  0.041168863186441806\n",
      "Diff RMSE:  0.04108546984721741\n",
      "adjusted R2: 1.0011941541508989\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.0008230675204355262\n",
      "MAE of hist average:  0.03240918888178644\n",
      "Diff MAE:  0.031586121361350916\n",
      "RMSE of regression:  5.277414957451501e-06\n",
      "RMSE of hist average:  0.04220732832902981\n",
      "Diff RMSE:  0.042202050914072364\n",
      "adjusted R2: 1.0001044666918713\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.0016090069364065366\n",
      "MAE of hist average:  0.03163422480557368\n",
      "Diff MAE:  0.030025217869167144\n",
      "RMSE of regression:  4.1608523985717024e-05\n",
      "RMSE of hist average:  0.04139534350998232\n",
      "Diff RMSE:  0.04135373498599661\n",
      "adjusted R2: 1.0006365585372665\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.0014383859324833862\n",
      "MAE of hist average:  0.03241842427786493\n",
      "Diff MAE:  0.03098003834538154\n",
      "RMSE of regression:  3.9525653534690576e-05\n",
      "RMSE of hist average:  0.04205029655890334\n",
      "Diff RMSE:  0.04201077090536865\n",
      "adjusted R2: 1.0007215786745804\n",
      "-------------------------------------------------------------------\n",
      "MAE of regression:  0.0010065365941326519\n",
      "MAE of hist average:  0.032353985198962695\n",
      "Diff MAE:  0.03134744860483004\n",
      "RMSE of regression:  1.9112193765569544e-05\n",
      "RMSE of hist average:  0.04217360939986197\n",
      "Diff RMSE:  0.0421544972060964\n",
      "adjusted R2: 1.0003685331841516\n",
      "-------------------------------------------------------------------\n",
      "the main take way: Mean(Diff_MAE) is 0.0309 Mean(Diff_RMSE) is 0.0419\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import GradientBoostingRegressor\n",
    "\n",
    "diff_MAE = []\n",
    "diff_RMSE = []\n",
    "\n",
    "for train_index, test_index in kf:\n",
    "    x_train, x_test = x.iloc[train_index], x.iloc[test_index]\n",
    "    ret_train, ret_test = ret.iloc[train_index], ret.iloc[test_index]\n",
    "    \n",
    "    regr = GradientBoostingRegressor(n_estimators=25, learning_rate=0.5, max_depth=25, random_state=0, loss='ls')\n",
    "    regr.fit(x_train, ret_train)\n",
    "    \n",
    "    ret_pred = regr.predict(x_test)\n",
    "    \n",
    "    diff = df['his_avg'].iloc[train_index] - df['premium'].iloc[train_index]\n",
    "    mae_avg = diff.abs().sum() / len(train_index)\n",
    "    diff_MAE.append(mae_avg - mean_absolute_error(ret_test, ret_pred))\n",
    "    rmse_avg = math.sqrt(np.power(diff, 2).sum() / len(train_index))\n",
    "    diff_RMSE.append(rmse_avg - mean_squared_error(ret_test, ret_pred))\n",
    "\n",
    "    r_squared = r2_score(ret_test, ret_pred)\n",
    "    adjusted_r_squared = 1 - (1-r_squared)*(len(ret_test)-1)/(len(ret_test)-x_test.shape[1]-1)\n",
    "    \n",
    "    print(\"MAE of regression: \", mean_absolute_error(ret_test, ret_pred))\n",
    "    print(\"MAE of hist average: \", mae_avg)\n",
    "    print(\"Diff MAE: \", mae_avg - mean_absolute_error(ret_test, ret_pred))\n",
    "    print(\"RMSE of regression: \", mean_squared_error(ret_test, ret_pred))\n",
    "    print(\"RMSE of hist average: \", rmse_avg)\n",
    "    print(\"Diff RMSE: \", rmse_avg - mean_squared_error(ret_test, ret_pred))\n",
    "    print(\"adjusted R2:\", adjusted_r_squared)\n",
    "    print(\"-------------------------------------------------------------------\")\n",
    "    \n",
    "print( \"the main take way: Mean(Diff_MAE) is\", round(sum(diff_MAE)/len(diff_MAE),4),\n",
    "      \"Mean(Diff_RMSE) is\", round(sum(diff_RMSE)/len(diff_RMSE),4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}