AFML 3.2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Triple Barrier Method\n",
    "\n",
    "This notebook will cover partial exercise answers:\n",
    "\n",
    "* Exercise 3.5\n",
    "\n",
    "As we go along, there will be some explanations.\n",
    "\n",
    "More importantly, this method can be applied not just within mean-reversion strategy but also other strategies as well. \n",
    "\n",
    "Most of the functions below can be found under research/Labels.\n",
    "\n",
    "Contact: boyboi86@gmail.com"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Num of CPU core:  4\n",
      "Machine info:  Windows-10-10.0.18362-SP0\n",
      "Python 3.7.4 (default, Aug  9 2019, 18:34:13) [MSC v.1915 64 bit (AMD64)]\n",
      "Numpy 1.17.3\n",
      "Pandas 1.0.3\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Wei_X\\Anaconda3\\lib\\site-packages\\statsmodels\\tools\\_testing.py:19: FutureWarning: pandas.util.testing is deprecated. Use the functions in the public API at pandas.testing instead.\n",
      "  import pandas.util.testing as tm\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 1500x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import research as rs\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "p = print\n",
    "\n",
    "#pls take note of version\n",
    "#numpy 1.17.3\n",
    "#pandas 1.0.3\n",
    "#sklearn 0.21.3\n",
    "\n",
    "dollar = pd.read_csv('./research/Sample_data/dollar_bars.txt', \n",
    "                 sep=',', \n",
    "                 header=0, \n",
    "                 parse_dates = True, \n",
    "                 index_col=['date_time'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def bband(data: pd.DataFrame, window: int = 21, width: float = 0.001):\n",
    "    avg = data['close'].ewm(span = window).mean()\n",
    "    std = avg * width\n",
    "    upper = avg + std\n",
    "    lower = avg - std\n",
    "    return avg, upper, lower, std\n",
    "\n",
    "dollar['ewm'], dollar['upper'], dollar['lower'], dollar['std'] = bband(dollar)\n",
    "\n",
    "# Check for normality, serial correlation, overall statistical properties, frequency count stability"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Num of times upper limit touched: 8495\n",
      "Num of times lower limit touched: 7811\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "-1.0    8495\n",
       " 1.0    7811\n",
       "Name: side, dtype: int64"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dollar['side'] = np.nan\n",
    "\n",
    "def side_pick(data: pd.DataFrame):\n",
    "    for i in np.arange(data.index.shape[0]):\n",
    "        if (data['close'].iloc[i] >= data['upper'].iloc[i]):\n",
    "            data['side'].iat[i] = -1\n",
    "        elif (data['close'].iloc[i] <= data['lower'].iloc[i]):\n",
    "            data['side'].iat[i] = 1\n",
    "    return data\n",
    "\n",
    "upper = dollar[dollar['upper'] < dollar['close']] # short signal\n",
    "lower = dollar[dollar['lower'] > dollar['close']] # long signal\n",
    "\n",
    "p(\"Num of times upper limit touched: {0}\\nNum of times lower limit touched: {1}\"\n",
    "  .format(upper.count()[0], \n",
    "          lower.count()[0]))\n",
    "\n",
    "# Recall white test as a benchmark and until this stage we filtered all those which did not meet min return\n",
    "dollar = side_pick(dollar)\n",
    "dollar.dropna(inplace= True)\n",
    "dollar['side'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>open</th>\n",
       "      <th>high</th>\n",
       "      <th>low</th>\n",
       "      <th>close</th>\n",
       "      <th>cum_vol</th>\n",
       "      <th>cum_dollar</th>\n",
       "      <th>cum_ticks</th>\n",
       "      <th>ewm</th>\n",
       "      <th>upper</th>\n",
       "      <th>lower</th>\n",
       "      <th>std</th>\n",
       "      <th>side</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>date_time</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2015-01-02 07:07:35.156</th>\n",
       "      <td>2056.75</td>\n",
       "      <td>2067.25</td>\n",
       "      <td>2056.25</td>\n",
       "      <td>2064.00</td>\n",
       "      <td>33968</td>\n",
       "      <td>70010061.25</td>\n",
       "      <td>14514</td>\n",
       "      <td>2060.547619</td>\n",
       "      <td>2062.608167</td>\n",
       "      <td>2058.487071</td>\n",
       "      <td>2.060548</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 14:19:33.847</th>\n",
       "      <td>2061.00</td>\n",
       "      <td>2064.25</td>\n",
       "      <td>2058.75</td>\n",
       "      <td>2063.75</td>\n",
       "      <td>33958</td>\n",
       "      <td>70000723.25</td>\n",
       "      <td>12332</td>\n",
       "      <td>2061.404789</td>\n",
       "      <td>2063.466193</td>\n",
       "      <td>2059.343384</td>\n",
       "      <td>2.061405</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 14:33:39.311</th>\n",
       "      <td>2063.75</td>\n",
       "      <td>2064.75</td>\n",
       "      <td>2060.00</td>\n",
       "      <td>2064.50</td>\n",
       "      <td>33944</td>\n",
       "      <td>70001009.00</td>\n",
       "      <td>12846</td>\n",
       "      <td>2062.050864</td>\n",
       "      <td>2064.112915</td>\n",
       "      <td>2059.988814</td>\n",
       "      <td>2.062051</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 14:42:28.315</th>\n",
       "      <td>2064.50</td>\n",
       "      <td>2066.50</td>\n",
       "      <td>2063.25</td>\n",
       "      <td>2066.00</td>\n",
       "      <td>33901</td>\n",
       "      <td>70010093.25</td>\n",
       "      <td>13032</td>\n",
       "      <td>2062.788295</td>\n",
       "      <td>2064.851084</td>\n",
       "      <td>2060.725507</td>\n",
       "      <td>2.062788</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 15:01:45.497</th>\n",
       "      <td>2063.50</td>\n",
       "      <td>2064.00</td>\n",
       "      <td>2058.75</td>\n",
       "      <td>2058.75</td>\n",
       "      <td>34008</td>\n",
       "      <td>70122046.75</td>\n",
       "      <td>13873</td>\n",
       "      <td>2062.252963</td>\n",
       "      <td>2064.315216</td>\n",
       "      <td>2060.190710</td>\n",
       "      <td>2.062253</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:22:33.456</th>\n",
       "      <td>2231.25</td>\n",
       "      <td>2233.75</td>\n",
       "      <td>2231.00</td>\n",
       "      <td>2233.50</td>\n",
       "      <td>31355</td>\n",
       "      <td>70001983.50</td>\n",
       "      <td>2453</td>\n",
       "      <td>2237.134042</td>\n",
       "      <td>2239.371176</td>\n",
       "      <td>2234.896908</td>\n",
       "      <td>2.237134</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:32:57.188</th>\n",
       "      <td>2233.50</td>\n",
       "      <td>2235.00</td>\n",
       "      <td>2232.50</td>\n",
       "      <td>2233.25</td>\n",
       "      <td>31349</td>\n",
       "      <td>70029302.00</td>\n",
       "      <td>2621</td>\n",
       "      <td>2236.780947</td>\n",
       "      <td>2239.017728</td>\n",
       "      <td>2234.544167</td>\n",
       "      <td>2.236781</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:44:21.481</th>\n",
       "      <td>2233.25</td>\n",
       "      <td>2234.00</td>\n",
       "      <td>2230.50</td>\n",
       "      <td>2230.75</td>\n",
       "      <td>31362</td>\n",
       "      <td>70017565.50</td>\n",
       "      <td>2836</td>\n",
       "      <td>2236.232680</td>\n",
       "      <td>2238.468912</td>\n",
       "      <td>2233.996447</td>\n",
       "      <td>2.236233</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:50:57.567</th>\n",
       "      <td>2230.75</td>\n",
       "      <td>2231.75</td>\n",
       "      <td>2229.25</td>\n",
       "      <td>2229.75</td>\n",
       "      <td>31386</td>\n",
       "      <td>70009159.50</td>\n",
       "      <td>2101</td>\n",
       "      <td>2235.643345</td>\n",
       "      <td>2237.878988</td>\n",
       "      <td>2233.407702</td>\n",
       "      <td>2.235643</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:55:33.160</th>\n",
       "      <td>2229.75</td>\n",
       "      <td>2231.25</td>\n",
       "      <td>2229.25</td>\n",
       "      <td>2231.00</td>\n",
       "      <td>31399</td>\n",
       "      <td>70020888.50</td>\n",
       "      <td>2017</td>\n",
       "      <td>2235.221223</td>\n",
       "      <td>2237.456444</td>\n",
       "      <td>2232.986002</td>\n",
       "      <td>2.235221</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>16306 rows × 12 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                            open     high      low    close  cum_vol  \\\n",
       "date_time                                                              \n",
       "2015-01-02 07:07:35.156  2056.75  2067.25  2056.25  2064.00    33968   \n",
       "2015-01-02 14:19:33.847  2061.00  2064.25  2058.75  2063.75    33958   \n",
       "2015-01-02 14:33:39.311  2063.75  2064.75  2060.00  2064.50    33944   \n",
       "2015-01-02 14:42:28.315  2064.50  2066.50  2063.25  2066.00    33901   \n",
       "2015-01-02 15:01:45.497  2063.50  2064.00  2058.75  2058.75    34008   \n",
       "...                          ...      ...      ...      ...      ...   \n",
       "2016-12-30 20:22:33.456  2231.25  2233.75  2231.00  2233.50    31355   \n",
       "2016-12-30 20:32:57.188  2233.50  2235.00  2232.50  2233.25    31349   \n",
       "2016-12-30 20:44:21.481  2233.25  2234.00  2230.50  2230.75    31362   \n",
       "2016-12-30 20:50:57.567  2230.75  2231.75  2229.25  2229.75    31386   \n",
       "2016-12-30 20:55:33.160  2229.75  2231.25  2229.25  2231.00    31399   \n",
       "\n",
       "                          cum_dollar  cum_ticks          ewm        upper  \\\n",
       "date_time                                                                   \n",
       "2015-01-02 07:07:35.156  70010061.25      14514  2060.547619  2062.608167   \n",
       "2015-01-02 14:19:33.847  70000723.25      12332  2061.404789  2063.466193   \n",
       "2015-01-02 14:33:39.311  70001009.00      12846  2062.050864  2064.112915   \n",
       "2015-01-02 14:42:28.315  70010093.25      13032  2062.788295  2064.851084   \n",
       "2015-01-02 15:01:45.497  70122046.75      13873  2062.252963  2064.315216   \n",
       "...                              ...        ...          ...          ...   \n",
       "2016-12-30 20:22:33.456  70001983.50       2453  2237.134042  2239.371176   \n",
       "2016-12-30 20:32:57.188  70029302.00       2621  2236.780947  2239.017728   \n",
       "2016-12-30 20:44:21.481  70017565.50       2836  2236.232680  2238.468912   \n",
       "2016-12-30 20:50:57.567  70009159.50       2101  2235.643345  2237.878988   \n",
       "2016-12-30 20:55:33.160  70020888.50       2017  2235.221223  2237.456444   \n",
       "\n",
       "                               lower       std  side  \n",
       "date_time                                             \n",
       "2015-01-02 07:07:35.156  2058.487071  2.060548  -1.0  \n",
       "2015-01-02 14:19:33.847  2059.343384  2.061405  -1.0  \n",
       "2015-01-02 14:33:39.311  2059.988814  2.062051  -1.0  \n",
       "2015-01-02 14:42:28.315  2060.725507  2.062788  -1.0  \n",
       "2015-01-02 15:01:45.497  2060.190710  2.062253   1.0  \n",
       "...                              ...       ...   ...  \n",
       "2016-12-30 20:22:33.456  2234.896908  2.237134   1.0  \n",
       "2016-12-30 20:32:57.188  2234.544167  2.236781   1.0  \n",
       "2016-12-30 20:44:21.481  2233.996447  2.236233   1.0  \n",
       "2016-12-30 20:50:57.567  2233.407702  2.235643   1.0  \n",
       "2016-12-30 20:55:33.160  2232.986002  2.235221   1.0  \n",
       "\n",
       "[16306 rows x 12 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "copy_dollar = dollar.copy() # make a back copy to be used in later exercise\n",
    "\n",
    "copy_dollar #up till this point the below dataframe should look like this, before tri_bar func. This is our primary model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "d_vol = rs.vol(dollar['close'], span0 = 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DatetimeIndex(['2015-01-02 14:19:33.847000', '2015-01-02 14:33:39.311000',\n",
       "               '2015-01-02 14:42:28.315000', '2015-01-02 15:01:45.497000',\n",
       "               '2015-01-02 15:22:54.187000', '2015-01-02 15:32:59.861000',\n",
       "               '2015-01-02 15:39:14.826000', '2015-01-02 15:43:25.099000',\n",
       "               '2015-01-02 15:48:54.420000', '2015-01-02 15:57:26.907000',\n",
       "               ...\n",
       "               '2016-12-30 19:02:57.783000', '2016-12-30 19:29:47.411000',\n",
       "               '2016-12-30 19:47:05.557000', '2016-12-30 19:55:31.030000',\n",
       "               '2016-12-30 20:12:10.314000', '2016-12-30 20:22:33.456000',\n",
       "               '2016-12-30 20:32:57.188000', '2016-12-30 20:44:21.481000',\n",
       "               '2016-12-30 20:50:57.567000', '2016-12-30 20:55:33.160000'],\n",
       "              dtype='datetime64[ns]', length=15459, freq=None)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "events = rs.cs_filter(dollar['close'], \n",
    "                    limit = d_vol.mean())\n",
    "\n",
    "events"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2015-01-02 14:19:33.847   2015-01-04 23:20:12.567\n",
       "2015-01-02 14:33:39.311   2015-01-04 23:20:12.567\n",
       "2015-01-02 14:42:28.315   2015-01-04 23:20:12.567\n",
       "2015-01-02 15:01:45.497   2015-01-04 23:20:12.567\n",
       "2015-01-02 15:22:54.187   2015-01-04 23:20:12.567\n",
       "                                    ...          \n",
       "2016-12-29 16:18:10.918   2016-12-30 17:17:03.543\n",
       "2016-12-29 16:43:29.395   2016-12-30 17:17:03.543\n",
       "2016-12-29 17:03:04.248   2016-12-30 17:17:03.543\n",
       "2016-12-29 17:17:04.110   2016-12-30 17:44:08.768\n",
       "2016-12-29 18:58:02.400   2016-12-30 19:02:57.783\n",
       "Name: date_time, Length: 15434, dtype: datetime64[ns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vb = rs.vert_barrier(data = dollar['close'], \n",
    "                 events = events, \n",
    "                 period = 'days', \n",
    "                 freq = 1)\n",
    "\n",
    "vb # Show some example output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Wei_X\\Desktop\\Python\\research\\Labels\\triple_barrier_method.py:75: UserWarning: Data and events index shape must be same, reindex data to fit events\n",
      "  warnings.warn('Data and events index shape must be same, reindex data to fit events')\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[                                             t1                      sl  pt\n",
      "2015-01-05 14:54:26.286 2015-01-06 15:16:04.445 2015-01-05 16:15:59.512 NaT\n",
      "2015-01-05 14:57:13.616 2015-01-06 15:16:04.445 2015-01-05 16:21:16.062 NaT\n",
      "2015-01-05 15:01:57.494 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "2015-01-05 15:07:29.012 2015-01-06 15:16:04.445 2015-01-05 17:26:59.327 NaT\n",
      "2015-01-05 15:13:09.655 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "...                                         ...                     ...  ..\n",
      "2015-09-09 08:33:28.634 2015-09-10 09:33:29.450                     NaT NaT\n",
      "2015-09-09 10:21:19.955 2015-09-10 11:43:14.539                     NaT NaT\n",
      "2015-09-09 12:02:05.852 2015-09-10 12:19:25.229                     NaT NaT\n",
      "2015-09-09 13:06:01.330 2015-09-10 13:30:20.359                     NaT NaT\n",
      "2015-09-09 13:31:24.524 2015-09-10 13:35:08.110                     NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns]] this out\n",
      "[                                             t1                      sl  pt\n",
      "2015-01-05 14:54:26.286 2015-01-06 15:16:04.445 2015-01-05 16:15:59.512 NaT\n",
      "2015-01-05 14:57:13.616 2015-01-06 15:16:04.445 2015-01-05 16:21:16.062 NaT\n",
      "2015-01-05 15:01:57.494 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "2015-01-05 15:07:29.012 2015-01-06 15:16:04.445 2015-01-05 17:26:59.327 NaT\n",
      "2015-01-05 15:13:09.655 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "...                                         ...                     ...  ..\n",
      "2015-09-09 08:33:28.634 2015-09-10 09:33:29.450                     NaT NaT\n",
      "2015-09-09 10:21:19.955 2015-09-10 11:43:14.539                     NaT NaT\n",
      "2015-09-09 12:02:05.852 2015-09-10 12:19:25.229                     NaT NaT\n",
      "2015-09-09 13:06:01.330 2015-09-10 13:30:20.359                     NaT NaT\n",
      "2015-09-09 13:31:24.524 2015-09-10 13:35:08.110                     NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns],                                              t1  sl  pt\n",
      "2015-09-09 13:36:53.472 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:40:56.133 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:45:08.189 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:51:11.450 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 14:02:04.210 2015-09-10 14:05:19.003 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-04-05 01:16:26.261 2016-04-06 02:45:07.823 NaT NaT\n",
      "2016-04-05 06:02:04.305 2016-04-06 07:07:24.337 NaT NaT\n",
      "2016-04-05 07:11:07.085 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 07:37:56.273 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 08:24:13.547 2016-04-06 09:05:34.702 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns]] this out\n",
      "[                                             t1                      sl  pt\n",
      "2015-01-05 14:54:26.286 2015-01-06 15:16:04.445 2015-01-05 16:15:59.512 NaT\n",
      "2015-01-05 14:57:13.616 2015-01-06 15:16:04.445 2015-01-05 16:21:16.062 NaT\n",
      "2015-01-05 15:01:57.494 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "2015-01-05 15:07:29.012 2015-01-06 15:16:04.445 2015-01-05 17:26:59.327 NaT\n",
      "2015-01-05 15:13:09.655 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "...                                         ...                     ...  ..\n",
      "2015-09-09 08:33:28.634 2015-09-10 09:33:29.450                     NaT NaT\n",
      "2015-09-09 10:21:19.955 2015-09-10 11:43:14.539                     NaT NaT\n",
      "2015-09-09 12:02:05.852 2015-09-10 12:19:25.229                     NaT NaT\n",
      "2015-09-09 13:06:01.330 2015-09-10 13:30:20.359                     NaT NaT\n",
      "2015-09-09 13:31:24.524 2015-09-10 13:35:08.110                     NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns],                                              t1  sl  pt\n",
      "2015-09-09 13:36:53.472 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:40:56.133 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:45:08.189 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:51:11.450 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 14:02:04.210 2015-09-10 14:05:19.003 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-04-05 01:16:26.261 2016-04-06 02:45:07.823 NaT NaT\n",
      "2016-04-05 06:02:04.305 2016-04-06 07:07:24.337 NaT NaT\n",
      "2016-04-05 07:11:07.085 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 07:37:56.273 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 08:24:13.547 2016-04-06 09:05:34.702 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns],                                              t1  sl  pt\n",
      "2016-04-05 09:17:57.890 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 10:11:56.123 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 11:26:40.687 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 12:10:26.367 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 12:55:36.469 2016-04-06 13:14:38.661 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-12-30 20:22:33.456                     NaT NaT NaT\n",
      "2016-12-30 20:32:57.188                     NaT NaT NaT\n",
      "2016-12-30 20:44:21.481                     NaT NaT NaT\n",
      "2016-12-30 20:50:57.567                     NaT NaT NaT\n",
      "2016-12-30 20:55:33.160                     NaT NaT NaT\n",
      "\n",
      "[5141 rows x 3 columns]] this out\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-05-24 19:47:10.399688 33.33% _pt_sl_t1 done after 0.38 minutes. Remaining 0.76 minutes.\r",
      "2020-05-24 19:47:10.399688 66.67% _pt_sl_t1 done after 0.38 minutes. Remaining 0.19 minutes.\r",
      "2020-05-24 19:47:10.493415 100.0% _pt_sl_t1 done after 0.38 minutes. Remaining 0.0 minutes.\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>t1</th>\n",
       "      <th>trgt</th>\n",
       "      <th>side</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2015-01-05 14:54:26.286</th>\n",
       "      <td>2015-01-05 16:15:59.512</td>\n",
       "      <td>0.002238</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 14:57:13.616</th>\n",
       "      <td>2015-01-05 16:21:16.062</td>\n",
       "      <td>0.002462</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 15:01:57.494</th>\n",
       "      <td>2015-01-05 17:32:22.888</td>\n",
       "      <td>0.002779</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 15:07:29.012</th>\n",
       "      <td>2015-01-05 17:26:59.327</td>\n",
       "      <td>0.002819</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 15:13:09.655</th>\n",
       "      <td>2015-01-05 17:32:22.888</td>\n",
       "      <td>0.002874</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 16:18:10.918</th>\n",
       "      <td>2016-12-30 17:17:03.543</td>\n",
       "      <td>0.004250</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 16:43:29.395</th>\n",
       "      <td>2016-12-30 17:17:03.543</td>\n",
       "      <td>0.004166</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 17:03:04.248</th>\n",
       "      <td>2016-12-30 17:17:03.543</td>\n",
       "      <td>0.004090</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 17:17:04.110</th>\n",
       "      <td>2016-12-30 17:44:08.768</td>\n",
       "      <td>0.004016</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 18:58:02.400</th>\n",
       "      <td>2016-12-30 19:02:57.783</td>\n",
       "      <td>0.003979</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>15400 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             t1      trgt  side\n",
       "2015-01-05 14:54:26.286 2015-01-05 16:15:59.512  0.002238   1.0\n",
       "2015-01-05 14:57:13.616 2015-01-05 16:21:16.062  0.002462   1.0\n",
       "2015-01-05 15:01:57.494 2015-01-05 17:32:22.888  0.002779   1.0\n",
       "2015-01-05 15:07:29.012 2015-01-05 17:26:59.327  0.002819   1.0\n",
       "2015-01-05 15:13:09.655 2015-01-05 17:32:22.888  0.002874   1.0\n",
       "...                                         ...       ...   ...\n",
       "2016-12-29 16:18:10.918 2016-12-30 17:17:03.543  0.004250   1.0\n",
       "2016-12-29 16:43:29.395 2016-12-30 17:17:03.543  0.004166   1.0\n",
       "2016-12-29 17:03:04.248 2016-12-30 17:17:03.543  0.004090   1.0\n",
       "2016-12-29 17:17:04.110 2016-12-30 17:44:08.768  0.004016   1.0\n",
       "2016-12-29 18:58:02.400 2016-12-30 19:02:57.783  0.003979   1.0\n",
       "\n",
       "[15400 rows x 3 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tb = rs.tri_barrier(data = dollar['close'], \n",
    "                events = events, \n",
    "                trgt = d_vol, \n",
    "                min_req = 0.002, \n",
    "                num_threads = 3, \n",
    "                ptSl = [0,2], #change ptSl into [0,2]\n",
    "                t1 = vb, \n",
    "                side = dollar['side'])\n",
    "\n",
    "tb # Show some example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ret</th>\n",
       "      <th>bin</th>\n",
       "      <th>side</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2015-01-05 14:54:26.286</th>\n",
       "      <td>-0.004556</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 14:57:13.616</th>\n",
       "      <td>-0.005051</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 15:01:57.494</th>\n",
       "      <td>-0.005798</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 15:07:29.012</th>\n",
       "      <td>-0.005790</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 15:13:09.655</th>\n",
       "      <td>-0.006656</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 16:18:10.918</th>\n",
       "      <td>-0.001672</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 16:43:29.395</th>\n",
       "      <td>-0.001894</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 17:03:04.248</th>\n",
       "      <td>-0.001226</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 17:17:04.110</th>\n",
       "      <td>-0.001450</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 18:58:02.400</th>\n",
       "      <td>-0.002899</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>15400 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                              ret  bin  side\n",
       "2015-01-05 14:54:26.286 -0.004556  0.0   1.0\n",
       "2015-01-05 14:57:13.616 -0.005051  0.0   1.0\n",
       "2015-01-05 15:01:57.494 -0.005798  0.0   1.0\n",
       "2015-01-05 15:07:29.012 -0.005790  0.0   1.0\n",
       "2015-01-05 15:13:09.655 -0.006656  0.0   1.0\n",
       "...                           ...  ...   ...\n",
       "2016-12-29 16:18:10.918 -0.001672  0.0   1.0\n",
       "2016-12-29 16:43:29.395 -0.001894  0.0   1.0\n",
       "2016-12-29 17:03:04.248 -0.001226  0.0   1.0\n",
       "2016-12-29 17:17:04.110 -0.001450  0.0   1.0\n",
       "2016-12-29 18:58:02.400 -0.002899  0.0   1.0\n",
       "\n",
       "[15400 rows x 3 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m_label = rs.meta_label(data = dollar['close'],\n",
    "                      events = tb,\n",
    "                      drop = False)\n",
    "\n",
    "m_label # Show some example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0    0.519545\n",
       "1.0    0.480455\n",
       "Name: bin, dtype: float64"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m_label['bin'].value_counts(normalize=True)\n",
    "\n",
    "# Here is a quick look at our 'bin' values.\n",
    "# Slight imbalanced sample, but not much harm\n",
    "# 51.95% of the sample based on parameter touched vertical barrier first"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Exercise 3.5b\n",
    "\n",
    "Here onwards we will be using sklearn modules to perform ML related task."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# A quick look at what we have till date using both primary and secondary model\n",
    "# as seen in previous example, only 48.04% was labeled 1. \n",
    "# Hence precision 1.0 = 0.48 (48% of the sample is relevant), while recall = 1 means fully correct (based on the 48% sample)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The below function report_matrix is what we have till date using both primary (bband func) and secondary model (tri_bar func).\n",
    "\n",
    "#### Classification Report\n",
    "\n",
    "As seen in previous example, only 48.0455% was labeled 1. \n",
    "\n",
    "Hence precision 1.0 = 0.48 (48.0455% of the sample is relevant). It's basically ML's way of saying are these \"features\" relevant when tested.\n",
    "\n",
    "While recall = 1 means fully correct (based on the 48% sample). In the case where ML model is fitted, this result will mean the percentage of \"correct\" label was chosen. In short, is the ML model reliability in True positive identification based on given sample.\n",
    "\n",
    "#### Confusion Matrix\n",
    "\n",
    "8001 = False Positive (51.95%)\n",
    "7399 = True Positive (48.0455%)\n",
    "\n",
    "#### Accuracy Score\n",
    "\n",
    "Is a mere reflection of True Positive, which again is 48.0455%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Classification Report\n",
      "=======================================================\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.00      0.00      0.00      8001\n",
      "         1.0       0.48      1.00      0.65      7399\n",
      "\n",
      "    accuracy                           0.48     15400\n",
      "   macro avg       0.24      0.50      0.32     15400\n",
      "weighted avg       0.23      0.48      0.31     15400\n",
      "\n",
      "\n",
      " Confusion Matrix\n",
      "=======================================================\n",
      "[[TN, FP]\n",
      " [FN, TP]]\n",
      "\n",
      "[[   0 8001]\n",
      " [   0 7399]]\n",
      "\n",
      " Accuracy Score\n",
      "=======================================================\n",
      "0.48045454545454547\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Wei_X\\Anaconda3\\lib\\site-packages\\sklearn\\metrics\\classification.py:1437: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n"
     ]
    }
   ],
   "source": [
    "# this func can be found under Tools/stats_rpt\n",
    "\n",
    "forecast = rs.report_matrix(actual_data = m_label, \n",
    "                            prediction_data = None,\n",
    "                            ROC = None)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Built a list of features.\n",
    "\n",
    "1. Volatility\n",
    "2. Autocorrelation\n",
    "3. Moving average\n",
    "4. log-price return (optional)\n",
    "5. Stationary series based on cumulative sum log-price return (optional)\n",
    "\n",
    "The last 2 items will be explained at AFML chapter 5, fractional differentiated features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>open</th>\n",
       "      <th>high</th>\n",
       "      <th>low</th>\n",
       "      <th>close</th>\n",
       "      <th>cum_vol</th>\n",
       "      <th>cum_dollar</th>\n",
       "      <th>cum_ticks</th>\n",
       "      <th>ewm</th>\n",
       "      <th>upper</th>\n",
       "      <th>lower</th>\n",
       "      <th>std</th>\n",
       "      <th>side</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>date_time</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2015-01-02 07:07:35.156</th>\n",
       "      <td>2056.75</td>\n",
       "      <td>2067.25</td>\n",
       "      <td>2056.25</td>\n",
       "      <td>2064.00</td>\n",
       "      <td>33968</td>\n",
       "      <td>70010061.25</td>\n",
       "      <td>14514</td>\n",
       "      <td>2060.547619</td>\n",
       "      <td>2062.608167</td>\n",
       "      <td>2058.487071</td>\n",
       "      <td>2.060548</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 14:19:33.847</th>\n",
       "      <td>2061.00</td>\n",
       "      <td>2064.25</td>\n",
       "      <td>2058.75</td>\n",
       "      <td>2063.75</td>\n",
       "      <td>33958</td>\n",
       "      <td>70000723.25</td>\n",
       "      <td>12332</td>\n",
       "      <td>2061.404789</td>\n",
       "      <td>2063.466193</td>\n",
       "      <td>2059.343384</td>\n",
       "      <td>2.061405</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 14:33:39.311</th>\n",
       "      <td>2063.75</td>\n",
       "      <td>2064.75</td>\n",
       "      <td>2060.00</td>\n",
       "      <td>2064.50</td>\n",
       "      <td>33944</td>\n",
       "      <td>70001009.00</td>\n",
       "      <td>12846</td>\n",
       "      <td>2062.050864</td>\n",
       "      <td>2064.112915</td>\n",
       "      <td>2059.988814</td>\n",
       "      <td>2.062051</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 14:42:28.315</th>\n",
       "      <td>2064.50</td>\n",
       "      <td>2066.50</td>\n",
       "      <td>2063.25</td>\n",
       "      <td>2066.00</td>\n",
       "      <td>33901</td>\n",
       "      <td>70010093.25</td>\n",
       "      <td>13032</td>\n",
       "      <td>2062.788295</td>\n",
       "      <td>2064.851084</td>\n",
       "      <td>2060.725507</td>\n",
       "      <td>2.062788</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 15:01:45.497</th>\n",
       "      <td>2063.50</td>\n",
       "      <td>2064.00</td>\n",
       "      <td>2058.75</td>\n",
       "      <td>2058.75</td>\n",
       "      <td>34008</td>\n",
       "      <td>70122046.75</td>\n",
       "      <td>13873</td>\n",
       "      <td>2062.252963</td>\n",
       "      <td>2064.315216</td>\n",
       "      <td>2060.190710</td>\n",
       "      <td>2.062253</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:22:33.456</th>\n",
       "      <td>2231.25</td>\n",
       "      <td>2233.75</td>\n",
       "      <td>2231.00</td>\n",
       "      <td>2233.50</td>\n",
       "      <td>31355</td>\n",
       "      <td>70001983.50</td>\n",
       "      <td>2453</td>\n",
       "      <td>2237.134042</td>\n",
       "      <td>2239.371176</td>\n",
       "      <td>2234.896908</td>\n",
       "      <td>2.237134</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:32:57.188</th>\n",
       "      <td>2233.50</td>\n",
       "      <td>2235.00</td>\n",
       "      <td>2232.50</td>\n",
       "      <td>2233.25</td>\n",
       "      <td>31349</td>\n",
       "      <td>70029302.00</td>\n",
       "      <td>2621</td>\n",
       "      <td>2236.780947</td>\n",
       "      <td>2239.017728</td>\n",
       "      <td>2234.544167</td>\n",
       "      <td>2.236781</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:44:21.481</th>\n",
       "      <td>2233.25</td>\n",
       "      <td>2234.00</td>\n",
       "      <td>2230.50</td>\n",
       "      <td>2230.75</td>\n",
       "      <td>31362</td>\n",
       "      <td>70017565.50</td>\n",
       "      <td>2836</td>\n",
       "      <td>2236.232680</td>\n",
       "      <td>2238.468912</td>\n",
       "      <td>2233.996447</td>\n",
       "      <td>2.236233</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:50:57.567</th>\n",
       "      <td>2230.75</td>\n",
       "      <td>2231.75</td>\n",
       "      <td>2229.25</td>\n",
       "      <td>2229.75</td>\n",
       "      <td>31386</td>\n",
       "      <td>70009159.50</td>\n",
       "      <td>2101</td>\n",
       "      <td>2235.643345</td>\n",
       "      <td>2237.878988</td>\n",
       "      <td>2233.407702</td>\n",
       "      <td>2.235643</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:55:33.160</th>\n",
       "      <td>2229.75</td>\n",
       "      <td>2231.25</td>\n",
       "      <td>2229.25</td>\n",
       "      <td>2231.00</td>\n",
       "      <td>31399</td>\n",
       "      <td>70020888.50</td>\n",
       "      <td>2017</td>\n",
       "      <td>2235.221223</td>\n",
       "      <td>2237.456444</td>\n",
       "      <td>2232.986002</td>\n",
       "      <td>2.235221</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>16306 rows × 12 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                            open     high      low    close  cum_vol  \\\n",
       "date_time                                                              \n",
       "2015-01-02 07:07:35.156  2056.75  2067.25  2056.25  2064.00    33968   \n",
       "2015-01-02 14:19:33.847  2061.00  2064.25  2058.75  2063.75    33958   \n",
       "2015-01-02 14:33:39.311  2063.75  2064.75  2060.00  2064.50    33944   \n",
       "2015-01-02 14:42:28.315  2064.50  2066.50  2063.25  2066.00    33901   \n",
       "2015-01-02 15:01:45.497  2063.50  2064.00  2058.75  2058.75    34008   \n",
       "...                          ...      ...      ...      ...      ...   \n",
       "2016-12-30 20:22:33.456  2231.25  2233.75  2231.00  2233.50    31355   \n",
       "2016-12-30 20:32:57.188  2233.50  2235.00  2232.50  2233.25    31349   \n",
       "2016-12-30 20:44:21.481  2233.25  2234.00  2230.50  2230.75    31362   \n",
       "2016-12-30 20:50:57.567  2230.75  2231.75  2229.25  2229.75    31386   \n",
       "2016-12-30 20:55:33.160  2229.75  2231.25  2229.25  2231.00    31399   \n",
       "\n",
       "                          cum_dollar  cum_ticks          ewm        upper  \\\n",
       "date_time                                                                   \n",
       "2015-01-02 07:07:35.156  70010061.25      14514  2060.547619  2062.608167   \n",
       "2015-01-02 14:19:33.847  70000723.25      12332  2061.404789  2063.466193   \n",
       "2015-01-02 14:33:39.311  70001009.00      12846  2062.050864  2064.112915   \n",
       "2015-01-02 14:42:28.315  70010093.25      13032  2062.788295  2064.851084   \n",
       "2015-01-02 15:01:45.497  70122046.75      13873  2062.252963  2064.315216   \n",
       "...                              ...        ...          ...          ...   \n",
       "2016-12-30 20:22:33.456  70001983.50       2453  2237.134042  2239.371176   \n",
       "2016-12-30 20:32:57.188  70029302.00       2621  2236.780947  2239.017728   \n",
       "2016-12-30 20:44:21.481  70017565.50       2836  2236.232680  2238.468912   \n",
       "2016-12-30 20:50:57.567  70009159.50       2101  2235.643345  2237.878988   \n",
       "2016-12-30 20:55:33.160  70020888.50       2017  2235.221223  2237.456444   \n",
       "\n",
       "                               lower       std  side  \n",
       "date_time                                             \n",
       "2015-01-02 07:07:35.156  2058.487071  2.060548  -1.0  \n",
       "2015-01-02 14:19:33.847  2059.343384  2.061405  -1.0  \n",
       "2015-01-02 14:33:39.311  2059.988814  2.062051  -1.0  \n",
       "2015-01-02 14:42:28.315  2060.725507  2.062788  -1.0  \n",
       "2015-01-02 15:01:45.497  2060.190710  2.062253   1.0  \n",
       "...                              ...       ...   ...  \n",
       "2016-12-30 20:22:33.456  2234.896908  2.237134   1.0  \n",
       "2016-12-30 20:32:57.188  2234.544167  2.236781   1.0  \n",
       "2016-12-30 20:44:21.481  2233.996447  2.236233   1.0  \n",
       "2016-12-30 20:50:57.567  2233.407702  2.235643   1.0  \n",
       "2016-12-30 20:55:33.160  2232.986002  2.235221   1.0  \n",
       "\n",
       "[16306 rows x 12 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Data that was copied earlier before tri_bar func, this is our primary model only\n",
    "\n",
    "copy_dollar # Show example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>close</th>\n",
       "      <th>ewm</th>\n",
       "      <th>upper</th>\n",
       "      <th>lower</th>\n",
       "      <th>std</th>\n",
       "      <th>side</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>date_time</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2015-01-02 07:07:35.156</th>\n",
       "      <td>2064.00</td>\n",
       "      <td>2060.547619</td>\n",
       "      <td>2062.608167</td>\n",
       "      <td>2058.487071</td>\n",
       "      <td>2.060548</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 14:19:33.847</th>\n",
       "      <td>2063.75</td>\n",
       "      <td>2061.404789</td>\n",
       "      <td>2063.466193</td>\n",
       "      <td>2059.343384</td>\n",
       "      <td>2.061405</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 14:33:39.311</th>\n",
       "      <td>2064.50</td>\n",
       "      <td>2062.050864</td>\n",
       "      <td>2064.112915</td>\n",
       "      <td>2059.988814</td>\n",
       "      <td>2.062051</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 14:42:28.315</th>\n",
       "      <td>2066.00</td>\n",
       "      <td>2062.788295</td>\n",
       "      <td>2064.851084</td>\n",
       "      <td>2060.725507</td>\n",
       "      <td>2.062788</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-02 15:01:45.497</th>\n",
       "      <td>2058.75</td>\n",
       "      <td>2062.252963</td>\n",
       "      <td>2064.315216</td>\n",
       "      <td>2060.190710</td>\n",
       "      <td>2.062253</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:22:33.456</th>\n",
       "      <td>2233.50</td>\n",
       "      <td>2237.134042</td>\n",
       "      <td>2239.371176</td>\n",
       "      <td>2234.896908</td>\n",
       "      <td>2.237134</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:32:57.188</th>\n",
       "      <td>2233.25</td>\n",
       "      <td>2236.780947</td>\n",
       "      <td>2239.017728</td>\n",
       "      <td>2234.544167</td>\n",
       "      <td>2.236781</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:44:21.481</th>\n",
       "      <td>2230.75</td>\n",
       "      <td>2236.232680</td>\n",
       "      <td>2238.468912</td>\n",
       "      <td>2233.996447</td>\n",
       "      <td>2.236233</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:50:57.567</th>\n",
       "      <td>2229.75</td>\n",
       "      <td>2235.643345</td>\n",
       "      <td>2237.878988</td>\n",
       "      <td>2233.407702</td>\n",
       "      <td>2.235643</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-30 20:55:33.160</th>\n",
       "      <td>2231.00</td>\n",
       "      <td>2235.221223</td>\n",
       "      <td>2237.456444</td>\n",
       "      <td>2232.986002</td>\n",
       "      <td>2.235221</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>16306 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                           close          ewm        upper        lower  \\\n",
       "date_time                                                                 \n",
       "2015-01-02 07:07:35.156  2064.00  2060.547619  2062.608167  2058.487071   \n",
       "2015-01-02 14:19:33.847  2063.75  2061.404789  2063.466193  2059.343384   \n",
       "2015-01-02 14:33:39.311  2064.50  2062.050864  2064.112915  2059.988814   \n",
       "2015-01-02 14:42:28.315  2066.00  2062.788295  2064.851084  2060.725507   \n",
       "2015-01-02 15:01:45.497  2058.75  2062.252963  2064.315216  2060.190710   \n",
       "...                          ...          ...          ...          ...   \n",
       "2016-12-30 20:22:33.456  2233.50  2237.134042  2239.371176  2234.896908   \n",
       "2016-12-30 20:32:57.188  2233.25  2236.780947  2239.017728  2234.544167   \n",
       "2016-12-30 20:44:21.481  2230.75  2236.232680  2238.468912  2233.996447   \n",
       "2016-12-30 20:50:57.567  2229.75  2235.643345  2237.878988  2233.407702   \n",
       "2016-12-30 20:55:33.160  2231.00  2235.221223  2237.456444  2232.986002   \n",
       "\n",
       "                              std  side  \n",
       "date_time                                \n",
       "2015-01-02 07:07:35.156  2.060548  -1.0  \n",
       "2015-01-02 14:19:33.847  2.061405  -1.0  \n",
       "2015-01-02 14:33:39.311  2.062051  -1.0  \n",
       "2015-01-02 14:42:28.315  2.062788  -1.0  \n",
       "2015-01-02 15:01:45.497  2.062253   1.0  \n",
       "...                           ...   ...  \n",
       "2016-12-30 20:22:33.456  2.237134   1.0  \n",
       "2016-12-30 20:32:57.188  2.236781   1.0  \n",
       "2016-12-30 20:44:21.481  2.236233   1.0  \n",
       "2016-12-30 20:50:57.567  2.235643   1.0  \n",
       "2016-12-30 20:55:33.160  2.235221   1.0  \n",
       "\n",
       "[16306 rows x 6 columns]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# drop redundant columns and keep crossing moving avaerages\n",
    "pri_dollar = copy_dollar.drop(['open', 'high', 'low', 'cum_vol', 'cum_dollar', 'cum_ticks'], axis = 1)\n",
    "\n",
    "#include volatility, autocorrelation\n",
    "pri_dollar"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "#include original volatility\n",
    "pri_dollar['volatility'] = rs.vol(pri_dollar.close, span0 = 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Wei_X\\Desktop\\Python\\research\\Features\\fractional_diff.py:93: UserWarning: thres val <= 1.e-2 may not suit non-trend series, may take up longer than expected to calculate val i.e. 1e-5\n",
      "  warnings.warn('thres val <= 1.e-2 may not suit non-trend series, may take up longer than expected to calculate val i.e. 1e-5')\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "ADF & KPSS: Weak evidence for stationary\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Wei_X\\Anaconda3\\lib\\site-packages\\statsmodels\\tsa\\stattools.py:1685: FutureWarning: The behavior of using lags=None will change in the next release. Currently lags=None is the same as lags='legacy', and so a sample-size lag length is used. After the next release, the default will change to be the same as lags='auto' which uses an automatic lag length selection method. To silence this warning, either use 'auto' or 'legacy'\n",
      "  warn(msg, FutureWarning)\n",
      "C:\\Users\\Wei_X\\Anaconda3\\lib\\site-packages\\statsmodels\\tsa\\stattools.py:1710: InterpolationWarning: p-value is greater than the indicated p-value\n",
      "  warn(\"p-value is greater than the indicated p-value\", InterpolationWarning)\n"
     ]
    }
   ],
   "source": [
    "# Optional: getting stationarity feature\n",
    "pri_dollar['log_price'] = pri_dollar.close.apply(np.log)\n",
    "pri_dollar['log_return'] = pri_dollar.log_price.diff()\n",
    "\n",
    "cs_log = pri_dollar.log_price.diff().dropna().to_frame()\n",
    "pri_dollar['stationary'] = rs.fracDiff_FFD(data = cs_log, d = 1.99999889 , thres = 1e-5)\n",
    "\n",
    "rs.unit_root(pri_dollar['stationary'].dropna()) #check for stationarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "pri_dollar.dropna(inplace = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Wei_X\\Anaconda3\\lib\\site-packages\\statsmodels\\tsa\\base\\tsa_model.py:219: ValueWarning: A date index has been provided, but it has no associated frequency information and so will be ignored when e.g. forecasting.\n",
      "  ' ignored when e.g. forecasting.', ValueWarning)\n",
      "C:\\Users\\Wei_X\\Anaconda3\\lib\\site-packages\\statsmodels\\tsa\\base\\tsa_model.py:219: ValueWarning: A date index has been provided, but it has no associated frequency information and so will be ignored when e.g. forecasting.\n",
      "  ' ignored when e.g. forecasting.', ValueWarning)\n",
      "C:\\Users\\Wei_X\\Anaconda3\\lib\\site-packages\\statsmodels\\base\\model.py:512: ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals\n",
      "  \"Check mle_retvals\", ConvergenceWarning)\n",
      "C:\\Users\\Wei_X\\Anaconda3\\lib\\site-packages\\statsmodels\\tsa\\base\\tsa_model.py:219: ValueWarning: A date index has been provided, but it has no associated frequency information and so will be ignored when e.g. forecasting.\n",
      "  ' ignored when e.g. forecasting.', ValueWarning)\n"
     ]
    }
   ],
   "source": [
    "# autocorrelation residual feature, we will add AR features up to 2 lags\n",
    "from statsmodels.tsa.arima_model import ARMA\n",
    "\n",
    "pri_dollar['ar_0'] = ARMA(pri_dollar['stationary'], order=(0,0)).fit().resid\n",
    "pri_dollar['ar_1'] = ARMA(pri_dollar['stationary'], order=(1,0)).fit().resid\n",
    "pri_dollar['ar_2'] = ARMA(pri_dollar['stationary'], order=(2,0)).fit().resid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "#final dataset\n",
    "\n",
    "secondary_dollar = pri_dollar.copy()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Note\n",
    "\n",
    "* Good to include volume based or volume-weighted indicator as a predictive feature i.e. OBV, VWAP"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Note**\n",
    "\n",
    "May try to add other types of trend related features as part of experimental Mathematics. (aka Trial & error)\n",
    "\n",
    "* Good to include volume based or volume-weighted indicator as a predictive feature i.e. OBV, VWAP\n",
    "* If not, try to add price based as predictive feature i.e. MOM, RSI"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Wei_X\\Desktop\\Python\\research\\Labels\\triple_barrier_method.py:75: UserWarning: Data and events index shape must be same, reindex data to fit events\n",
      "  warnings.warn('Data and events index shape must be same, reindex data to fit events')\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[                                             t1  sl  pt\n",
      "2015-09-09 13:36:53.472 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:40:56.133 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:45:08.189 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:51:11.450 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 14:02:04.210 2015-09-10 14:05:19.003 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-04-05 01:16:26.261 2016-04-06 02:45:07.823 NaT NaT\n",
      "2016-04-05 06:02:04.305 2016-04-06 07:07:24.337 NaT NaT\n",
      "2016-04-05 07:11:07.085 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 07:37:56.273 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 08:24:13.547 2016-04-06 09:05:34.702 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns]] this out\n",
      "[                                             t1  sl  pt\n",
      "2015-09-09 13:36:53.472 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:40:56.133 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:45:08.189 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:51:11.450 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 14:02:04.210 2015-09-10 14:05:19.003 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-04-05 01:16:26.261 2016-04-06 02:45:07.823 NaT NaT\n",
      "2016-04-05 06:02:04.305 2016-04-06 07:07:24.337 NaT NaT\n",
      "2016-04-05 07:11:07.085 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 07:37:56.273 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 08:24:13.547 2016-04-06 09:05:34.702 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns],                                              t1                      sl  pt\n",
      "2015-01-05 14:54:26.286 2015-01-06 15:16:04.445 2015-01-05 16:15:59.512 NaT\n",
      "2015-01-05 14:57:13.616 2015-01-06 15:16:04.445 2015-01-05 16:21:16.062 NaT\n",
      "2015-01-05 15:01:57.494 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "2015-01-05 15:07:29.012 2015-01-06 15:16:04.445 2015-01-05 17:26:59.327 NaT\n",
      "2015-01-05 15:13:09.655 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "...                                         ...                     ...  ..\n",
      "2015-09-09 08:33:28.634 2015-09-10 09:33:29.450                     NaT NaT\n",
      "2015-09-09 10:21:19.955 2015-09-10 11:43:14.539                     NaT NaT\n",
      "2015-09-09 12:02:05.852 2015-09-10 12:19:25.229                     NaT NaT\n",
      "2015-09-09 13:06:01.330 2015-09-10 13:30:20.359                     NaT NaT\n",
      "2015-09-09 13:31:24.524 2015-09-10 13:35:08.110                     NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns]] this out\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-05-24 19:47:44.210234 33.33% _pt_sl_t1 done after 0.38 minutes. Remaining 0.75 minutes.\r",
      "2020-05-24 19:47:44.303960 66.67% _pt_sl_t1 done after 0.38 minutes. Remaining 0.19 minutes.\r"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[                                             t1  sl  pt\n",
      "2015-09-09 13:36:53.472 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:40:56.133 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:45:08.189 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:51:11.450 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 14:02:04.210 2015-09-10 14:05:19.003 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-04-05 01:16:26.261 2016-04-06 02:45:07.823 NaT NaT\n",
      "2016-04-05 06:02:04.305 2016-04-06 07:07:24.337 NaT NaT\n",
      "2016-04-05 07:11:07.085 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 07:37:56.273 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 08:24:13.547 2016-04-06 09:05:34.702 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns],                                              t1                      sl  pt\n",
      "2015-01-05 14:54:26.286 2015-01-06 15:16:04.445 2015-01-05 16:15:59.512 NaT\n",
      "2015-01-05 14:57:13.616 2015-01-06 15:16:04.445 2015-01-05 16:21:16.062 NaT\n",
      "2015-01-05 15:01:57.494 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "2015-01-05 15:07:29.012 2015-01-06 15:16:04.445 2015-01-05 17:26:59.327 NaT\n",
      "2015-01-05 15:13:09.655 2015-01-06 15:16:04.445 2015-01-05 17:32:22.888 NaT\n",
      "...                                         ...                     ...  ..\n",
      "2015-09-09 08:33:28.634 2015-09-10 09:33:29.450                     NaT NaT\n",
      "2015-09-09 10:21:19.955 2015-09-10 11:43:14.539                     NaT NaT\n",
      "2015-09-09 12:02:05.852 2015-09-10 12:19:25.229                     NaT NaT\n",
      "2015-09-09 13:06:01.330 2015-09-10 13:30:20.359                     NaT NaT\n",
      "2015-09-09 13:31:24.524 2015-09-10 13:35:08.110                     NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns],                                              t1  sl  pt\n",
      "2016-04-05 09:17:57.890 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 10:11:56.123 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 11:26:40.687 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 12:10:26.367 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 12:55:36.469 2016-04-06 13:14:38.661 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-12-30 20:22:33.456                     NaT NaT NaT\n",
      "2016-12-30 20:32:57.188                     NaT NaT NaT\n",
      "2016-12-30 20:44:21.481                     NaT NaT NaT\n",
      "2016-12-30 20:50:57.567                     NaT NaT NaT\n",
      "2016-12-30 20:55:33.160                     NaT NaT NaT\n",
      "\n",
      "[5141 rows x 3 columns]] this out\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-05-24 19:47:44.834118 100.0% _pt_sl_t1 done after 0.39 minutes. Remaining 0.0 minutes.\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>ret</th>\n",
       "      <th>bin</th>\n",
       "      <th>side</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2015-01-05 14:54:26.286</th>\n",
       "      <td>-0.004556</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 14:57:13.616</th>\n",
       "      <td>-0.005051</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 15:01:57.494</th>\n",
       "      <td>-0.005798</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 15:07:29.012</th>\n",
       "      <td>-0.005790</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2015-01-05 15:13:09.655</th>\n",
       "      <td>-0.006656</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 16:18:10.918</th>\n",
       "      <td>-0.001672</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 16:43:29.395</th>\n",
       "      <td>-0.001894</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 17:03:04.248</th>\n",
       "      <td>-0.001226</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 17:17:04.110</th>\n",
       "      <td>-0.001450</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-12-29 18:58:02.400</th>\n",
       "      <td>-0.002899</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>15400 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                              ret  bin  side\n",
       "2015-01-05 14:54:26.286 -0.004556  0.0   1.0\n",
       "2015-01-05 14:57:13.616 -0.005051  0.0   1.0\n",
       "2015-01-05 15:01:57.494 -0.005798  0.0   1.0\n",
       "2015-01-05 15:07:29.012 -0.005790  0.0   1.0\n",
       "2015-01-05 15:13:09.655 -0.006656  0.0   1.0\n",
       "...                           ...  ...   ...\n",
       "2016-12-29 16:18:10.918 -0.001672  0.0   1.0\n",
       "2016-12-29 16:43:29.395 -0.001894  0.0   1.0\n",
       "2016-12-29 17:03:04.248 -0.001226  0.0   1.0\n",
       "2016-12-29 17:17:04.110 -0.001450  0.0   1.0\n",
       "2016-12-29 18:58:02.400 -0.002899  0.0   1.0\n",
       "\n",
       "[15400 rows x 3 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Now we run all the steps to complete labels, to train random forest.\n",
    "# we will use both primary & secondary model\n",
    "\n",
    "events0 = rs.cs_filter(secondary_dollar['close'], \n",
    "                    limit = secondary_dollar['volatility'].mean())\n",
    "\n",
    "vb0 = rs.vert_barrier(data = secondary_dollar['close'], \n",
    "                 events = events0, \n",
    "                 period = 'days', \n",
    "                 freq = 1)\n",
    "\n",
    "tb0 = rs.tri_barrier(data = secondary_dollar['close'], \n",
    "                events = events0, \n",
    "                trgt = secondary_dollar['volatility'], \n",
    "                min_req = 0.002, \n",
    "                num_threads = 3, \n",
    "                ptSl = [0,2], #change ptSl into [0,2]\n",
    "                t1 = vb0, \n",
    "                side = secondary_dollar['side'])\n",
    "\n",
    "m_label0 = rs.meta_label(data = secondary_dollar['close'],\n",
    "                      events = tb0,\n",
    "                      drop = 0.05)\n",
    "\n",
    "m_label0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0    8001\n",
       "1.0    7399\n",
       "Name: bin, dtype: int64"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m_label0['bin'].value_counts() \n",
    "\n",
    "# we still get back the same count. This is correct. \n",
    "# Tri_bar func is to calculate if vert_bar was triggered and consolidates the target.\n",
    "# while label will check which are the ones that hitted vertical barriers or non-profitable will be label 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='entropy',\n",
       "                       max_depth=7, max_features='auto', max_leaf_nodes=None,\n",
       "                       min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                       min_samples_leaf=1, min_samples_split=2,\n",
       "                       min_weight_fraction_leaf=0.0, n_estimators=500,\n",
       "                       n_jobs=None, oob_score=False, random_state=42, verbose=0,\n",
       "                       warm_start=False)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# At this stage you may wish to run Grid search CV, but I'm skipping that.\n",
    "\n",
    "n_estimators, max_depth, c_random_state = 500, 7, 42\n",
    "\n",
    "# Random Forest Model\n",
    "rf = RandomForestClassifier(max_depth=max_depth, \n",
    "                            n_estimators=n_estimators,\n",
    "                            criterion='entropy', \n",
    "                            class_weight = None, #This will be cover in next few chapters\n",
    "                            random_state=c_random_state)\n",
    "\n",
    "X = secondary_dollar.reindex(m_label0.index) # this dataframe only contain all our features\n",
    "y = m_label0['bin']\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, shuffle=False)\n",
    "\n",
    "rf.fit(X_train, y_train.values.ravel())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Matrix training report for primary model & secondary model\n",
      "\n",
      " Classification Report\n",
      "=======================================================\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.64      0.83      0.72      5721\n",
      "         1.0       0.71      0.48      0.57      5059\n",
      "\n",
      "    accuracy                           0.66     10780\n",
      "   macro avg       0.67      0.65      0.65     10780\n",
      "weighted avg       0.67      0.66      0.65     10780\n",
      "\n",
      "\n",
      " Confusion Matrix\n",
      "=======================================================\n",
      "[[TN, FP]\n",
      " [FN, TP]]\n",
      "\n",
      "[[4733  988]\n",
      " [2655 2404]]\n",
      "\n",
      " Accuracy Score\n",
      "=======================================================\n",
      "0.6620593692022263\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Performance Metrics\n",
    "y_prob = rf.predict_proba(X_train)[:, 1] #here we are only interested in True positive\n",
    "y_pred = rf.predict(X_train)\n",
    "\n",
    "p('Matrix training report for primary model & secondary model\\n')\n",
    "\n",
    "rs.report_matrix(actual_data = y_train, # we need to use our train data from train_test_split\n",
    "                 prediction_data = y_pred, \n",
    "                 ROC = y_prob)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Matrix test report for primary model & secondary model\n",
      "\n",
      " Classification Report\n",
      "=======================================================\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.54      0.37      0.43      2280\n",
      "         1.0       0.53      0.69      0.60      2340\n",
      "\n",
      "    accuracy                           0.53      4620\n",
      "   macro avg       0.53      0.53      0.52      4620\n",
      "weighted avg       0.53      0.53      0.52      4620\n",
      "\n",
      "\n",
      " Confusion Matrix\n",
      "=======================================================\n",
      "[[TN, FP]\n",
      " [FN, TP]]\n",
      "\n",
      "[[ 833 1447]\n",
      " [ 720 1620]]\n",
      "\n",
      " Accuracy Score\n",
      "=======================================================\n",
      "0.530952380952381\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEWCAYAAABrDZDcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nO3deZxN9f/A8dfbGERjZMtuxjp2Mlki+56tTVqUiFBa/OqbUkIlKi1EsiUiUomkJCnZd5J1MBj7vg+zvH9/3OsazHCNuXNn7n0/H4/7cD7nfM4572OY9z2fzzmfj6gqxhhj/FcGbwdgjDHGuywRGGOMn7NEYIwxfs4SgTHG+DlLBMYY4+csERhjjJ+zRGCMMX7OEoHxOSISKSLnReSMiBwQkfEicvtVde4RkT9F5LSInBSRn0Wk7FV1sovIpyKy23msCGc5d+pekTGeZYnA+KpWqno7UBmoArx+aYOI1AR+B2YABYBQYB2wSESKOetkAuYB5YBmQHbgHuAoUM1TQYtIRk8d25ikWCIwPk1VDwBzcCSESz4AJqjqZ6p6WlWPqeqbwFKgn7POk0AR4H5V3aiq8ap6SFXfUdXZiZ1LRMqJyFwROSYiB0XkDef68SLyboJ69UQkKkE5UkReE5H1wFkReVNEvr/q2J+JyFDncrCIjBWR/SKyV0TeFZGAW/yrMn7MEoHxaSJSCGgORDjLWXF8s5+WSPXvgMbO5UbAb6p6xs3zBAF/AL/huMsogeOOwl2PAvcBOYCJQAsRye48dgDQDpjsrPs1EOs8RxWgCfDMTZzLmCtYIjC+6icROQ3sAQ4BbzvX58Tx735/IvvsBy61/+dKok5SWgIHVHWIqkY77zSW3cT+Q1V1j6qeV9VdwGqgrXNbA+Ccqi4VkTtxJLaXVPWsqh4CPgHa38S5jLmCJQLjq9qqahBQDwjj8i/440A8kD+RffIDR5zLR5Ook5TCwPZkReqw56ryZBx3CQCPcfluoCgQCOwXkRMicgL4Esh7C+c2fs4SgfFpqvo3MB74yFk+CywBHk6kejsuN+f8ATQVkWxunmoPUDyJbWeBrAnK+RIL9aryNKCes2nrfi4ngj3ABSC3quZwfrKrajk34zTmGpYIjD/4FGgsIpc6jHsDT4nICyISJCJ3ODtzawL9nXUm4vil+4OIhIlIBhHJJSJviEiLRM4xC8gnIi+JSGbncas7t63F0eafU0TyAS/dKGBVPQz8BXwF7FTVTc71+3E88TTE+XhrBhEpLiJ1k/H3YgxgicD4Aecv1QnAW87yQqAp8ACOfoBdODpda6vqNmedCzg6jDcDc4FTwHIcTUzXtP2r6mkcHc2tgAPANqC+c/NEHI+nRuL4JT7VzdAnO2OYfNX6J4FMwEYcTV3fc3PNWMZcQWxiGmOM8W92R2CMMX7OEoExxvg5SwTGGOPnLBEYY4yfS3cDXOXOnVtDQkK8HYYxxqQrq1atOqKqeRLblu4SQUhICCtXrvR2GMYYk66IyK6ktlnTkDHG+DlLBMYY4+csERhjjJ9Ld30EiYmJiSEqKoro6Ghvh2JuUpYsWShUqBCBgYHeDsUYv+UTiSAqKoqgoCBCQkIQEW+HY9ykqhw9epSoqChCQ0O9HY4xfstjTUMiMk5EDonIhiS2i4gMdU4Ivl5E7kruuaKjo8mVK5clgXRGRMiVK5fdyRnjZZ7sIxiPY9LvpDQHSjo/XYEvbuVklgTSJ/u5GeN9HmsaUtUFIhJynSptcEwgrsBSEckhIvmd460bY4zfm7V+H1sPnOZiTAznzp3jwRqlqFQ4R4qfx5t9BAW5cnq+KOe6axKBiHTFcddAkSJFUiW4mxUQEECFChWIjY0lNDSUiRMnkiPHrf/AIiMjadmyJRs2JNrClmz9+vVj9OjR5MnjeNGwWbNmDBo0KEXPccnatWvZt28fLVokNp+LMeZqB05G0+yzBZw4F+NYofEAlCyU1+cSQWJtAolOjqCqo4BRAOHh4WlyAoXbbruNtWvXAvDUU08xfPhw+vTp4+Woru/ll1/mlVdeuen94uLiCAgIcLv+2rVrWblypSUCYxKx4/AZ/t56mMXbj3I6OoalO45dsX3fuJ4UDQ5gzJgx1K0Z4pEYvPkeQRSOCb8vKQTs81IsKapmzZrs3bsXgDNnztCwYUPuuusuKlSowIwZMwDHN/0yZcrQpUsXypUrR5MmTTh//jwAq1atolKlStSsWZPhw4e7jhsdHc3TTz9NhQoVqFKlCvPnzwdg/PjxtG3bllatWhEaGsrnn3/Oxx9/TJUqVahRowbHjh3DXfPmzaNKlSpUqFCBTp06ceHCBcAxtMeAAQOoXbs206ZNY/v27TRr1oyqVaty7733snnzZgCmTZtG+fLlqVSpEnXq1OHixYv07duXqVOnUrlyZaZOdXdyLmN8l6oy5p8dFHv9FxoM+Zv+P29k7saDrNl9gmJ5slEuf3YyrZjAng9b8/LTD7N+/Xrq1vXcbKTevCOYCTwvIlOA6sDJlOgf6P/zf2zcd+qWg0uobIHsvN3KvbnB4+LimDdvHp07dwYcz8lPnz6d7Nmzc+TIEWrUqEHr1q0B2LZtG99++y2jR4+mXbt2/PDDDzzxxBM8/fTTDBs2jLp16/Lqq6+6jn0pKfz7779s3ryZJk2asHXrVgA2bNjAmjVriI6OpkSJEgwePJg1a9bw8ssvM2HCBF566dppcj/55BO++eYbAAYPHkzdunXp2LEj8+bNo1SpUjz55JN88cUXrn2zZMnCwoULAWjYsCEjR46kZMmSLFu2jB49evDnn38yYMAA5syZQ8GCBTlx4gSZMmViwIABrFy5ks8//zw5f/3G+Iy4eGXG2r30+m7dFes/f6wK9UvnJfrMSXLmzImIML3IEQoPfpXw8HCPx+WxRCAi3wL1gNwiEgW8DQQCqOpIYDbQAogAzgFPeyqW1HD+/HkqV65MZGQkVatWpXHjxoAj87/xxhssWLCADBkysHfvXg4ePAhAaGgolSs75lOvWrUqkZGRnDx5khMnTriyf4cOHfj1118BWLhwIT179gQgLCyMokWLuhJB/fr1CQoKIigoiODgYFq1agVAhQoVWL9+faIxX900tG7dOkJDQylVqhRwuYnrUiJ45JFHAMddzuLFi3n44Ydd+166c6hVqxYdO3akXbt2PPDAA7f0d2pMehcfryzbeYy5Gw+yPPIoG/Ze/pKaM1smlr3RkMCADKgqkyZN4sUXX2TQoEF06dKF+++/P9Xi9ORTQ4/eYLsCz6X0ed395p7SLvURnDx5kpYtWzJ8+HBeeOEFJk2axOHDh1m1ahWBgYGEhIS4npvPnDmza/+AgADOnz+Pqib5SOX15pdOeKwMGTK4yhkyZCA2Ntata7jR/NXZsmUDID4+nhw5crj6RBIaOXIky5Yt45dffqFy5cqJ1jHG18XFK9PX7OWVaVd+829RIR9BmQPpVDuU0vmCANizZw/dunVj9uzZ1KhRg1q1aqV6vDbWUAoLDg5m6NChfPTRR8TExHDy5Eny5s1LYGAg8+fPZ9euJEeCBSBHjhwEBwe7mmAmTZrk2lanTh1XeevWrezevZvSpUunWOxhYWFERkYSEREBwMSJExNtl8yePTuhoaFMmzYNcCSQdesc/+C3b99O9erVGTBgALlz52bPnj0EBQVx+vTpFIvTmLTq0OloOoxdRvE3Zl+RBH55oTbr+jZhxONVGfxQRVcS+PbbbylXrhx//fUXn376KQsXLqRs2bKpHrdPDDGR1lSpUoVKlSoxZcoUHn/8cVq1akV4eDiVK1cmLCzshvt/9dVXdOrUiaxZs9K0aVPX+h49etCtWzcqVKhAxowZGT9+/BV3ArcqS5YsfPXVVzz88MPExsZy9913061bt0TrTpo0ie7du/Puu+8SExND+/btqVSpEq+++irbtm1DVWnYsCGVKlWiSJEiDBo0iMqVK/P666+7mpiMSc/i45UdR84w+98DxMTFM3x+BPEJbqpL3Xk7U7vW5I5smZI8xh133EH16tUZNWqUV4dZkRs1B6Q14eHhevXENJs2baJMmTJeisjcKvv5mfQk8shZ3pqxgX+2HblifaaMGbgtMIC3W5WlTeWCBGS4tok3NjaWTz75hIsXL7oeL79ec3BKEpFVqppoz7PdERhjzA2oKkt2HOWbpbuY/e8B1/qWFfNzX4X8NC57JxkDrt/Svm7dOjp37syqVato166dKwGkhWFWLBEYY8x1fPz7Fob+GXHFul6NS/F8/RJkSORb/9UuXLjAu+++y6BBg8iZMyfTpk3jwQcfTBMJ4BKfSQSpdXtlUlZ6a5o0/uHndfuYsmI3iyKOutY1DMvL6y3KUCLv7Td1rG3btjF48GAee+wxPv74Y3LlypXS4d4yn0gEWbJk4ejRozYUdTpzaT6CLFmyeDsUYwDHY5/F35jtKhfPk43bswTy+aNVKJwzq9vHOXPmDDNmzODxxx+nfPnybN68mWLFinki5BThE4mgUKFCREVFcfjwYW+HYm7SpRnKjPGmf6NOsi7qBG/+dHlwx687VaNuqTw3fay5c+fStWtXdu3axV133UWZMmXSdBIAH0kEgYGBNsOVMeamnYqO4YERi4k4dOaK9dsHtkj0qZ/rOX78OK+88grjxo2jVKlS/P333+nmaTifSATGGHOzJiyJpO+M/1zlsU+FU7lwDnLdfvPv5sTFxVGrVi22bt3K66+/Tt++fdNVk6clAmOMX4mNi6fs23O4GOsY4792idxM7FwtWf2LR44cIWfOnAQEBDBw4ECKFCnCXXcle9Zdr7EhJowxfuPzP7dRos+vriSw4NX6fPNM9ZtOAqrKhAkTKFWqFGPGjAGgbdu26TIJgN0RGGP8gKrSZcJK/th0CHAM//Dri3Vuuh8AYNeuXTz77LPMmTOHe+65hzp16qR0uKnOEoExxqetjzpB688Xucq/vFCbcgWCk3Wsb775hu7du6OqDBs2jB49epAhQ/pvWLFEYIzxSdExcTT7dAGRR8+51m1+pxlZAt2fZvVqefLkoVatWnz55ZcULVo0JcJMEywRGGN8yslzMXy5YDsj/truWjegTTmeTMZ8vzExMQwZMoSYmBjeeustmjZtSpMmTXzuxVVLBMaYdE9VmbYqilELdlzxTsB9FfPz+aNVkvWLe82aNXTu3Jk1a9bQvn37NDVIXEqzRGCMSbdORcfwxo//Mmv95enORaBn/RJ0qh1KjqxJzwWQlOjoaAYMGMAHH3xA7ty5+eGHH3x+2lVLBMaYdEdVeWHKWn5et8+17p7iufisfRXyBN3aZE0RERF89NFHPPnkkwwZMoQ77rjjVsNN8ywRGGPSlcXbj/DY6GWu8hM1ivBGizJkzZT8X2dnzpxh+vTpdOjQgfLly7Nlyxa/GrbGEoExJl348u/tfLUokgOnogHIlz0L8/6vLtky39qvsTlz5tC1a1f27NlDeHg4ZcqU8askAJYIjDFp3NkLsZR7e46rXCx3Nl5rHkbTcvlu6bhHjx6lV69eTJgwgbCwMP755590M0hcSrNEYIxJk35Zv58P5mxmV4L3AH7ofg9Vi956m/2lQeIiIiLo06cPb775ZroaJC6lWSIwxqQp0TFxDJ23zfUeQLE82bivQn5eaFiSwBvMC3wjhw8fJleuXAQEBDB48GCKFi1K5cqVUyLsdM0SgTEmTYiPV576ajn/bDviWvfJI5W4v8qtT1ykqowfP55evXoxaNAgnn32Wdq0aXPLx/UVlgiMMV6359g57v1gvqvc5d5QHqtelNDc2W752JGRkXTt2pW5c+dy7733Ur9+/Vs+pq+xRGCM8ZoNe0/y6R/b+GPTQQAyBWRgRZ9GBGcNTJHjT5w4ke7duyMijBgxgmeffdYnBolLaZYIjDGpbtzCnQyYtfGKdc/VL86rTcNS9Dx33nknderUYeTIkRQpUiRFj+1LLBEYY1LNsHnbGDJ3q6tcrkB2BrQpnyJPAoFjkLgPPviAuLg4+vbtS5MmTWjSpEmKHNuXWSIwxnjcwVPRdP56BRv2ngLg9swZ+fXFeymcM2uKnWP16tV06tSJdevW8dhjj7kGiTM3ZonAGONR0TFxVB84z1X+quPd1A/Lm2LHP3/+PP379+ejjz4iT548TJ8+nbZt26bY8f2BR3tNRKSZiGwRkQgR6Z3I9iIiMl9E1ojIehFp4cl4jDGp60JsHGX6/uYqRw66L0WTAMCOHTv4+OOP6dixIxs3brQkkAweSwQiEgAMB5oDZYFHRaTsVdXeBL5T1SpAe2CEp+IxxqSuDXtPUvrN31B1lHe+n3Lf806dOsX48eMBKFeuHNu2bWPMmDF+MVKoJ3jyjqAaEKGqO1T1IjAFuPoNDgWyO5eDgX0YY3zCwyOXANCoTF62vts8xdrrZ8+eTfny5encuTObNm0C8KlpI73Bk4mgILAnQTnKuS6hfsATIhIFzAZ6JnYgEekqIitFZOXhw4c9EasxJgWdPBfD+Zg4AMY8dTeZMt76r5ojR47QoUMH7rvvPoKCgli0aJHfDhKX0jyZCBJL/3pV+VFgvKoWAloAE0XkmphUdZSqhqtqeJ48eTwQqjEmpRw8FU2lAb8D0OwWRwi95NIgcVOmTKFv376sXr2aGjVqpMixjWefGooCCicoF+Lapp/OQDMAVV0iIlmA3MAhD8ZljPGQeZsO0vnrlYDjHYEvnrjrlo538OBB8uTJQ0BAAB999BFFixalYsWKKRGqScCTdwQrgJIiEioimXB0Bs+8qs5uoCGAiJQBsgDW9mNMOjRs3jZXEqhYKJhZPWsnu19AVRk7diylS5dm1KhRALRq1cqSgId47I5AVWNF5HlgDhAAjFPV/0RkALBSVWcC/weMFpGXcTQbdVTVq5uPjDFpXK/v1vLj6r0AfPBQRdqFF77BHknbsWMHXbp04c8//6Ru3bo0atQopcI0SfDoC2WqOhtHJ3DCdX0TLG8EankyBmOMZzX6+G8iDp0BYOQTVWlWPvn9Al9//TU9evQgICCAkSNH0qVLFxskLhXYm8XGmGSJj1dqD/6TfScdcwjP+7+6FM9z+y0ds0CBAjRo0IAvvviCQoVufR4C4x5LBMaYm3bmQizlE8wj/Of/1aVYMpLAxYsXGTRoEPHx8fTr14/GjRvTuHHjlAzVuMESgTHGbSsijzFp6S5+Wnv5AcC1fRuTI2ummz/WihV06tSJDRs20KFDBxskzossERhj3PLWTxuYuHSXq/xIeGEGPVjhpn95nzt3jr59+/LJJ5+QP39+Zs6cSatWrVI6XHMTLBEYY65r1a5jPDVuBWcuxAIwsXM1apfInexv7zt37mTYsGF06dKFwYMHExwcnJLhmmSwRGCMSdK2g6d58AvHmEHF82Tjk0cqU7FQjps+zsmTJ/nxxx95+umnKVeuHBERERQunPxHTE3KskRgjLnGishj9Ji0msOnLwBQMu/tzO1VN1nH+uWXX3j22WfZv38/NWvWJCwszJJAGmOJwBjjcvzsRZ6bvJrF248CkDljBoa0q0TLigVu+liHDx/mpZdeYvLkyZQvX54ff/yRsLCUnZPYpAxLBMYYAKat3MOr3693lQc/WIFH7k7ehO9xcXHUrl2bnTt30r9/f3r37k2mTDf/ZJFJHZYIjPFzB09F0+bzRRw45XgxLPftmVnRp2GyOoMPHDhA3rx5CQgIYMiQIYSEhFC+fPmUDtmkMHt32xg/tnr3caoPnMeBU9FkCczAT8/VYuWbjW46CcTHx/Pll19SqlQpvvzySwBatmxpSSCduOEdgYjcBrwEFFXVbiJSAiipqr96PDpjjMf0mLSK2f8eAKBVpQIMe7RKso4TERFBly5d+Ouvv2jQoAFNmzZNyTBNKnCnaWgc8C9Q21neB0wDLBEYkw5diI2j9JuXJ5Qf+mgVWle6+c5ggK+++ooePXqQKVMmRo8eTefOne3t4HTInURQUlUfFZGHAVT1nNhP2ph0KS5er0gCS19vSL7gLMk+XpEiRWjatCnDhw+nYMGrZ6I16YU7ieCic+YwBRCRUOCiR6MyxqS4fSfOc8+gP13lyEH33fQxLly4wPvvv098fDwDBgygYcOGNGzYMCXDNF7gTmfxO8BvQCER+RqYD7zh0aiMMSlq7Z4TVySBze80u+ljLFu2jKpVq9K/f392796NzSHlO26YCJydwg8DXYDpQDVV/cPTgRljUsap6BjaDl8EwAN3FSRy0H1kCQxwe/+zZ8/Sq1cvatasycmTJ5k1axbjx4+3vgAfcsNEICK/q+phVZ2hqj+p6iER+T01gjPG3JpZ6/dRsZ/jv2v90nn4uF3lmz7Grl27GDFiBN26deO///7jvvtuvknJpG1J9hE4J5zPAtwpIkHApfSfHUje64bGmFRx5MwF6n4wn7MX4wCoXDgHY5+62+39T5w4wffff88zzzxD2bJliYiIsBnDfNj1OoufA3oBeYH/uJwITgEjPRyXMSaZhs3bxpC5W13lH7rXpGrRnG7vP2PGDLp3786hQ4eoXbs2YWFhlgR8XJKJQFU/AT4RkZdU9dNUjMkYkwyqSo3353HwlGPE0C73htLnvrJu73/o0CFeeOEFpk6dSsWKFZk5c6YNEucnbvj4qKp+KiJhQFkcTUWX1k/2ZGDGmJvT89s1riQw47laVCrs/rwBcXFx1KpVi927d/Puu+/yv//9j8DAQE+FatIYd4aYeBNoAoQBc4CmwELAEoExaURsXDyz1u8H4I9edSiRN8it/fbt20e+fPkICAjgs88+IyQkhLJl3b+LML7BnfcIHgHqA/tVtQNQCRu11Jg049d/91Oij2PElweqFHQrCcTHx/PFF18QFhbGyJGOLr8WLVpYEvBT7vxCP6+qcSIS63x66ABQzMNxGWNuID5eeWnqWmau2wc4ho8e+ECFG+63detWunTpwoIFC2jUqBHNmzf3dKgmjXMnEawRkRw4Bp9bieOpodUejcoYc12xcfGEvfUbsfGOt3u/e7Ym1UJv/GTQ2LFjef7558mSJQvjxo2jY8eO9mKYuX4icA4u109VTwDDRWQOkF1VLREY4yUXY+Mp9eblwX/Xvd2E4Nvc69gNCQmhefPmDB8+nPz583sqRJPOXDcRqKqKyCygqrMckSpRGWMSpao8Onqpq7z13eZkyph0V9+FCxd45513AHj33XdtkDiTKHc6i5eLyF0ej8QYc10rI48R+vpsVu06DsDGAU2vmwQWL15M5cqVee+999i/f78NEmeS5E4iqI0jGWwRkdUiskZErGnImFQ0d+NBHhq5BICCOW5jzkt1yJop8Rv6M2fO8OKLL1K7dm3OnTvHb7/9xtixY60vwCTJnc7itsk9uIg0Az4DAoAxqjookTrtgH445jtYp6qPJfd8xviihduO0GXCSgD6tSpLx1qh162/e/duvvzyS5577jkGDhxIUJB77xQY/+XOm8Xbk3NgEQkAhgONgShghYjMVNWNCeqUBF4HaqnqcRHJm5xzGePLnhi7DIDezcOSTALHjx9n2rRpdO3albJly7Jjxw4KFEje9JPG/7jTNJRc1YAIVd2hqheBKUCbq+p0AYar6nEAVT3kwXiMSXcmLdsFQOk7g+hWt3iidaZPn07ZsmXp0aMHW7ZsAbAkYG6KJxNBQWBPgnKUc11CpYBSIrJIRJY6m5KuISJdRWSliKw8fPiwh8I1Jm1ZvvMYfaZvAGBsx/Brth84cICHH36YBx54gHz58rF8+XJKly6d2mEaH+DWUBEiUgjHJPbzRSQzkFFVz95ot0TWXf3YQkagJFAPKAT8IyLlne8tXN5JdRQwCiA8PNwefTA+771fNjL6n52A426g0B1Zr9geFxfHvffey549exg4cCCvvPKKDRJnks2dQec6Ac8DwUBxoCgwAmh0g12jgMIJyoWAfYnUWaqqMcBOEdmCIzGscCt6Y3zM3I0HefOnf12jiF79xnBUVBQFChQgICCAoUOHEhoaakNFm1vmTtPQC0ANHENLoKpbcUxWcyMrgJIiEuqc7aw9MPOqOj/hGNAOEcmNo6loh3uhG+MbVJXVu49z93t/0GXCSlcSmNi5misJxMfHM2zYMMLCwvjiiy8AaN68uSUBkyLcaRqKVtWLl55Bdj4NdMMHklU1VkSexzF0dQAwTlX/E5EBwEpVnenc1kRENgJxwKuqejSZ12JMuvPN0l28+dOGK9bN6lmb8gWDXeXNmzfzzDPPsGjRIpo2bUrLli1TO0zj49xJBItE5H9AFhGpj2MKy1nuHFxVZwOzr1rXN8Gy4pgOs5fbERvjA+LilcoDfud0dCwA1UNz0vGeEOqVzsttmQJc9caMGcPzzz9P1qxZ+frrr+nQoYO9GGZSnDuJ4H9AV2Az8CKOb/FfejIoY3zZhr0naTlsoas8pWsNahTLlWjd4sWL06pVKz7//HPuvPPO1ArR+Bm50fgjItIK+M3Zoet14eHhunLlSm+HYUyyPDd5Nb84ZxILyCBsGtDsivGCoqOjGTBgAAADBw70SozGN4nIKlW99jlk3OssbgdEiMhXItLU2UdgjLlJn/6x1ZUEhj5ahe0DW1yRBBYtWkTlypV5//33OXz4sA0SZ1LNDROBc3rKUsDPQCdgh4iM9HRgxviSv7Yc4tM/tgEw6ZnqtK50+c3f06dP07NnT+69914uXLjAnDlzGD16tPUFmFTj1gtlqnpBRGYA53E8AdQO6ObJwIzxBTPW7mXI71vZfewcAAPvr0CtErmvqBMVFcWYMWPo2bMn7733Hrfffrs3QjV+zJ0XyhrheAegEbAImADYCKHGXMeRMxd48IvF7Dp6zrXu/Qcq8Gi1IgAcPXqU7777ju7du1OmTBl27NhhM4YZr3HnjqAbjgHjeqrqeQ/HY0y6duhUNO/+ssk1oTzA0tcbki84C+B4eeyHH37gueee49ixYzRo0IDSpUtbEjBe5c4w1A+lRiDGpHdx8Uq1gfNc5UqFgvnpuVqutv79+/fz3HPPMX36dKpWrcrvv/9ug8SZNCHJRCAif6tqXRE5zpWDxQmOd8FyJrGrMX7l3MVYPvhtC+MXR7rW7Xy/xRWdvZcGidu7dy8ffPABL7/8MhkzutVFZ8GRaDAAABsvSURBVIzHXe9fYn3nn7mvU8cYv1ex3+/Exju+K+XLnoUlrzdwJYE9e/ZQsGBBAgICGD58OKGhoZQqVcqb4RpzjSQfH1XVeOfiWFWNS/gBxqZOeMakXYdPXyCk9y+uJLDz/RYsfaMhIkJcXBxDhw69YpC4pk2bWhIwaZI796YVExacL5Td7ZlwjEn7VJUvF+xg0K+bXev+6FXHdRewadMmOnfuzJIlS2jevDmtWrXyVqjGuOV6fQSvAb2BIBE5dmk1jv4CuyMwfuvJccv5Z9sRAMLyBfHbS3Vc20aNGkXPnj0JCgpi4sSJPP744/ZimEnzrndH8AEwBHgfR0IAwNk0ZIzfiYmLp2SfX13l5W80JG/2LFfUKVmyJPfffz9Dhw4lb153pu0wxvuSHHROREqq6jYRqZjYdlVd79HIkmCDzpnUpqr0+m4d09fsda1b1LsBBXPcxvnz5+nXrx8iwqBBg7wYpTHXd71B5653R9Ab6AwMT2SbAnUSWW+Mz2n48d/sOOyYortCwWB+6H4PmTJmYMGCBTzzzDNs27aNbt26oarWDGTSpSQTgap2dv55b+qFY0zaEXHoNI0+XuAqr36rMTmzZeLUqVO81Ls3X3zxBcWKFWPevHk0aNDAi5Eac2tuOPqoiDwgIkHO5d4i8p2IVPJ8aMZ4z0dztlyRBBa8Wp+c2TIBsG/fPsaPH0+vXr1Yv369JQGT7rkzH0E/VT0tIvcArYCp2Axlxoc1/+wfPp8fAcBbLcsSOeg+suo5RowYAUBYWBg7d+5kyJAhZMuWzZuhGpMi3EkEl54SagmMUNUfgMyeC8kY74g6fo6Q3r+waf8pAOa/Uo9OtUKYOnUqZcuW5aWXXmLr1q0ANm2k8SnuJIL9IjIcx1DUs0Ukk5v7GZNubD14mtqD57vK6/o2IfPFk7Rt25b27dtTtGhRVq1aZW8GG5/kzpvF7YAWwDBVPS4iBUjwXoEx6d2CrYd5ctxyADreE0K/1uWIi4ujap067N27l48++ogXX3zRBokzPsudYajPiMhGoJ6I1AP+UdVfb7CbMWle3xkbWBhxxPVo6H0V8vN0pduJi4sjICCAESNGUKxYMUqUKOHlSI3xLHeeGnoe+A4o4vx8JyI9PB2YMZ5y/mIc3b9ZxYQlu9hx+CzVQnPSv1VZih/8izJlyrgGiWvSpIklAeMX3LnX7QpUU9UzACIyEFgMjPBkYMZ4wnu/bGT0Pztd5e+71STLmX107tye5cuX07JlS9q2bevFCI1Jfe4kAgFiEpRjnOuMSTfi4pXib8x2lRuG5WXEE3fx1ZjRvPDCCwQHBzN58mTat29vbwcbv+NOIpgILBWRH3AkgLbA1x6NypgUljAJ/NGrLsXzZENEKFOmDA8//DCffvopefLk8WKExniPO53FH4jIfODSUBPdVHWFZ8MyJuWcir58Q/vvm/Xo3+9tAgICGDx4MHXr1qVu3bpejM4Y73P3fYALzs9555/GpAs7j5ylYr/fAWhcNCNVKldiyJAhnDlzhqRG3jXG39zwjkBE+gCPAdNxNA1NFpFJqvq+p4MzJrnG/LODsQt3sv9kNABZ4s4xpkc7ihcvzp9//kn9+vVvcARj/Ic7fQRPAFVV9RyAiLwHrMIxYY0xac68TQd595dNAFQPzUmDIoH0eqgur7zyCv379ydr1qxejtCYtMWdRLDrqnoZgR3uHFxEmgGfAQHAGFVNdOYOEXkImAbcrao264xJtlnr9/H85DUANAw+zNhn7wPggchI6ww2Jgnu9BGcA/4TkTEiMhr4FzghIh+LyMdJ7eSc5H440BwoCzwqImUTqRcEvAAsS84FGHPJ0TMXXEkgJmIxE9/u6hokzpKAMUlz547gF+fnkqVuHrsaEKGqOwBEZArQBth4Vb13cMyP/IqbxzXmGqpK1Xf/AODclsUU2z+P39essUHijHGDO4+Pjk3msQsCexKUo4DqCSuISBWgsKrOEpEkE4GIdMXxhjNFihRJZjjGVyWcVD7u7HFer5OHF15YREBAgJcjMyZ98ORwiom9nul6Xk9EMgCfAB1vdCBVHQWMAsfk9SkUn0nnzl+Mo+PohSzbc8a1bm7PGoSVsvGBjLkZnkwEUUDhBOVCwL4E5SCgPPCX85X+fMBMEWltHcbmRi5ejKFM399d5dolcjO2YziZM9pdgDE3y+0JZkTkZmclWwGUFJFQ52Q27YGZlzaq6klVza2qIaoagqPvwZKAuaH169dT5dHLLYmLelbmm2eqWxIwJpncGYa6moj8C2xzliuJyLAb7aeqscDzwBxgE/Cdqv4nIgNEpPUtxm381IgRI7j73oacLdkEgCW9G1CwYEEvR2VM+uZO09BQHPMV/wSgqutExK3XMlV1NjD7qnV9k6hbz51jGv+kqogIp+4oTf7u413r8+e4zXtBGeMj3EkEGVR111VD88YlVdmYlHT27FnefPNNzmcMYkX2ezl69iIAj1Uvwntty3s5OmN8gzuJYI+IVAPU+ZJYT2CrZ8MyBubNm0eXLl04mKkAedq+DmcvkitbJj5rX4XaJXN7OzxjfIY7ncXdgV44pqk8CNRwrjPGI06cOMEzzzxDo0aNyJgxoyMJABULBbPqrcaWBIxJYe68UHYIxxM/xqSKgwcPMmXKFF577TWy1nyccUt206TsnYx6MtzboRnjk9wZhno0CV4Eu0RVu3okIuOXLv3yf/HFFyldujSRkZGc5jbqf/QXAG+1vGaYKmNMCnGnj+CPBMtZgPu5cugIY5JNVZk0aRIvvvgiZ86coUWLFuQtGEL/uXv4eZ3j/cPqoTkpnNOGjjbGU9xpGpqasCwiE4G5HovI+I3du3fTrVs3fv31V2rWrMnYsWPZcj4bjQdcfmP4yZpF6deqnBejNMb3JWeIiVCgaEoHYvxLbGws9erV49ChQwwdOpQePXowYNYmvl7iGEa64z0hvNYsjNsy2dvCxniaO30Ex7ncR5ABOAb09mRQxnft2LGDokWLkjFjRkaPHk3x4sXJnqcAxfv85qozoVM16pSy+QOMSS3XfXxUHG+RVQLyOD93qGoxVf0uNYIzviM2NpbBgwdTtmxZhg8fDoAULE+H73Zw1zuXWxonP1PdkoAxqey6dwSqqiIyXVWrplZAxvesXbuWzp07s3r1au6//37qNmtDSO/Lcx0F3xbIa83CeKy6zTVhjDe400ewXETuUtXVHo/G+JzPP/+cl19+mVy5ctF75I/MP5qdNuM2uLZP6VqDGsVyeTFCY0ySiUBEMjpHEK0NdBGR7cBZHBPOqKrelUoxmnTo0iBxFStW5PHHHyci7Am+3XkBiKZIzqw0KnMnfVvZuwHGpAXXuyNYDtwFtE2lWIwPOHPmDH369CEwMJB33x/MZgqysEA7Yk9cAGDha/UpdIe9E2BMWnK9RCAAqro9lWIx6dzvv/9O165d2b13H6V6fcv3b11+EqhorqxM7FTdkoAxadD1EkEeEemV1EZV/dgD8Zh06Pjx47zcqxczIoXgNgMpclsw0c5tPRuU4OlaoeTMlsmrMRpjkna9RBAA3E7ik9Ab43Lo0CHmZa1LcHXHY58VCgbTpOyddKlTjCyB9kKYMWnd9RLBflUdkGqRmHTlwIEDfPvtt1Rv3YG3fzlAQJAjCazt25gcWe3bvzHpyQ37CIxJSFWZMGECL7/8Mhpak+CDpVzbvulc3ZKAMenQ9RJBw1SLwqQLkZGRPPvss/y9cS/5uk50rR/8YAUeudteBjMmvUoyEajqsdQMxKRtsbGx1LvvAbTpG+SrcnnC+Lkv16HknUFejMwYc6uSM/qo8SMREREULRrC+n2noNU7CJD79sx88kglahXPTYYM1oJoTHpnicAkKiYmhg8//JD+77xH/hcvjzHYrFw+RnawoaeM8SWWCMw1Vq9eTadne7AzQ8ErksDYp8JpEJbXi5EZYzzBEoG5wmefDaX/T2vI3vAt7nCuuzvkDqZ0rUmANQMZ45MsERgA/t5yiDELd/LP/uJkr14cgEerFaZHvRI2X7AxPs4SgZ87eeoUrftPYlfg5cc/W1UqwNO1QriryB3X2dMY4yssEfix/uNn89VmBWcSGNq+Mi0rFrAngYzxM5YI/FDk3oM0+XQBFwMcTT75s2Xg91cbEZQl0MuRGWO8wRKBnzl0Opp6w1aCMwmM6VCFRuUKeDkqY4w3eTQRiEgz4DMcI5mOUdVBV23vBTwDxAKHgU6qusuTMfmrqQs3MejXTRyPywxA+fy3M6NnHXsSyBjjuUQgIgHAcKAxEAWsEJGZqroxQbU1QLiqnhOR7sAHwCOeiskfxcbFU++dGURFZwIyc+ftGWlSviAD2pRDxJKAMcazdwTVgAhV3QEgIlOANoArEajq/AT1lwJPeDAev3IxNp7/fbuUn/47DjhGBB3YtACP1a/i3cCMMWmOJxNBQWBPgnIUUP069TsDvya2QUS6Al0BihSxUS6vR1V5bvJqZv97wLWuQKYLzH+rNZkDrUvIGHMtT/5mSKzdQROtKPIEEA7UTWy7qo4CRgGEh4cnegwDv6zfz3OTV7vKbYoH0rNRaUqEFvViVMaYtM6TiSAKKJygXAjYd3UlEWkE9AHqquoFD8bj096e8S9fL9ntKi/q3YCCOW67zh7GGOOQwYPHXgGUFJFQEckEtAdmJqwgIlWAL4HWqnrIg7H4tHkLl7uSQOlDf7G8192WBIwxbvPYHYGqxorI88AcHI+PjlPV/0RkALBSVWcCHwK3A9OcT7DsVtXWnorJFw36ZBgjDxYDIDxPPN8P+tDLERlj0huP9h6q6mxg9lXr+iZYbuTJ8/syVaXLhJX84UwCANN6tfRiRMaY9MoeI0lnTp06RffX32VR0OV+9VeblqZHveL2XoAxJlksEaQTqsqk6bN5czngTALlCmTn607VyH17Zu8GZ4xJ1ywRpAODfl7HyEVRV6z78KGKPBxeOIk9jDHGfZYI0rD4eKX96KUs33kMgKxcZGC7qrSsVIiMAZ584MsY408sEaRRQ2atYdjCy69dfN+5CuElbZRQY0zKs0SQxsTFxdPkvelsP5cFgDZlc/DhYzXJlNHuAIwxnmGJIA2ZuXgDL8zcBTiSQO+6+ejWvKp3gzLG+DxLBGmAqrJu9zFnEoDcgTEs7tuSTDZInDEmFdhvGi8bPH0ZXyw74irnvT2Q5W/e58WIjDH+xhKBFxw7e5FJS3Yydt6/nFDHmEANwvLSo15xwkNyejk6Y4y/sUSQyr74azuDf9vsLN1GhvgY3mldhsdrl/ZqXMYY/2WJIBW998tGRv+zE4D4bQsY+tS9tG3d1stRGWP8nSUCD4uJi2fUgh3sPX6OycsdE7aFn1nC2HFvEBwc7OXojDHGEoFHRcfEEfbWb65yUOaM/K95GB1qWGewMSbtsETgIXHxekUSeEj/4cN+79sIocaYNMcSgQdMXLCJt2bvcJW/eyAP1aoN8mJExhiTNEsEKWjhtiOM+CuCxduPAnBnhtPM7dOG7Nls2khjTNpliSAFTFy6i/4z/yM2XgEIDBDaVryTDx+xvgBjTNpnieAWbNh7kk7jV3Do9AUAYvZtZkinRrSvV8nLkRljjPssESTTql3HePCLJa5y/q0/MumjPhQrVuw6exljTNpjYxsnw8lzMa4kcOG/efQtdZDFP4yxJGCMSZfsjuAmbdq0ifsnRwJQt3Ag77/wCgUK2IQxxpj0yxKBmy5cuMDAgQMZe6QYGYNyA/D1c028HJUxxtw6axpyw9KlS6nQ+hm+vlDNlQSWvN7Ay1EZY0zKsDuCGxgyZAjDDodBlfYAVC6cg4mdqxGUJdDLkRljTMqwRJCE+Ph4YuLVkQSc/nqlHiG5s3kxKmOMSXmWCK5y4sQJnn+1D4eDSrAlUynX+tVvNSZntkxejMwYYzzDEkECU374idcWxyG5WrjWlc2fnVk9a5Mhgw0WZ4zxTZYIgEOHDnFfv0kczl4KCXS0/X/ySCUahN1J8G3WF2CM8W2WCIBZ6/ZyOLujGeiR8EIMfsiGiDDG+A+/TQS7d+9m4sSJbMjTkCU7HKOFTn6mOveUyO3lyIwxJnV59D0CEWkmIltEJEJEeieyPbOITHVuXyYiIZ6MBxxPA40YMYIKtRoz+nRlVxIY//TdlgSMMX7JY4lARAKA4UBzoCzwqIiUvapaZ+C4qpYAPgEGeyoegC1btlDtwWd5d0Usdzz+sWv9ot4NqFc6rydPbYwxaZYn7wiqARGqukNVLwJTgDZX1WkDfO1c/h5oKB6ay/HPTQdo+lUER0q3JdOdxQHoUa84kYPuo2AOmzjGGOO/PNlHUBDYk6AcBVRPqo6qxorISSAXcCRhJRHpCnQFKFKkSLKCiY2H3LcJjcLy8FrLStxh7wQYYwzg2USQ2Dd7TUYdVHUUMAogPDz8mu3uaFIuH03KtbhxRWOM8TOebBqKAgonKBcC9iVVR0QyAsHAMQ/GZIwx5iqeTAQrgJIiEioimYD2wMyr6swEnnIuPwT8qarJ+sZvjDEmeTzWNORs838emAMEAONU9T8RGQCsVNWZwFhgoohE4LgTaO+peIwxxiTOoy+UqepsYPZV6/omWI4GHvZkDMYYY67PJqYxxhg/Z4nAGGP8nCUCY4zxc5YIjDHGz0l6e1pTRA4Du5K5e26uemvZD9g1+we7Zv9wK9dcVFXzJLYh3SWCWyEiK1U13NtxpCa7Zv9g1+wfPHXN1jRkjDF+zhKBMcb4OX9LBKO8HYAX2DX7B7tm/+CRa/arPgJjjDHX8rc7AmOMMVexRGCMMX7OJxOBiDQTkS0iEiEivRPZnllEpjq3LxORkNSPMmW5cc29RGSjiKwXkXkiUtQbcaakG11zgnoPiYiKSLp/1NCdaxaRds6f9X8iMjm1Y0xpbvzbLiIi80VkjfPfd7qegUpExonIIRHZkMR2EZGhzr+P9SJy1y2fVFV96oNjyOvtQDEgE7AOKHtVnR7ASOdye2Cqt+NOhWuuD2R1Lnf3h2t21gsCFgBLgXBvx50KP+eSwBrgDmc5r7fjToVrHgV0dy6XBSK9HfctXnMd4C5gQxLbWwC/4pjhsQaw7FbP6Yt3BNWACFXdoaoXgSlAm6vqtAG+di5/DzQUkcSmzUwvbnjNqjpfVc85i0txzBiXnrnzcwZ4B/gAiE7N4DzEnWvuAgxX1eMAqnoolWNMae5cswLZncvBXDsTYrqiqgu4/kyNbYAJ6rAUyCEi+W/lnL6YCAoCexKUo5zrEq2jqrHASSBXqkTnGe5cc0KdcXyjSM9ueM0iUgUorKqzUjMwD3Ln51wKKCUii0RkqYg0S7XoPMOda+4HPCEiUTjmP+mZOqF5zc3+f78hj05M4yWJfbO/+hlZd+qkJ25fj4g8AYQDdT0akedd95pFJAPwCdAxtQJKBe78nDPiaB6qh+Ou7x8RKa+qJzwcm6e4c82PAuNVdYiI1MQx62F5VY33fHhekeK/v3zxjiAKKJygXIhrbxVddUQkI47byevdiqV17lwzItII6AO0VtULqRSbp9zomoOA8sBfIhKJoy11ZjrvMHb33/YMVY1R1Z3AFhyJIb1y55o7A98BqOoSIAuOwdl8lVv/32+GLyaCFUBJEQkVkUw4OoNnXlVnJvCUc/kh4E919sKkUze8ZmczyZc4kkB6bzeGG1yzqp5U1dyqGqKqITj6RVqr6krvhJsi3Pm3/ROOBwMQkdw4mop2pGqUKcuda94NNAQQkTI4EsHhVI0ydc0EnnQ+PVQDOKmq+2/lgD7XNKSqsSLyPDAHxxMH41T1PxEZAKxU1ZnAWBy3jxE47gTaey/iW+fmNX8I3A5Mc/aL71bV1l4L+ha5ec0+xc1rngM0EZGNQBzwqqoe9V7Ut8bNa/4/YLSIvIyjiaRjev5iJyLf4mjay+3s93gbCARQ1ZE4+kFaABHAOeDpWz5nOv77MsYYkwJ8sWnIGGPMTbBEYIwxfs4SgTHG+DlLBMYY4+csERhjjJ+zRGDSLBGJE5G1CT4h16kbktRojalNRMJFZKhzuZ6I3JNgWzcReTIVY6mc3kfjNJ7nc+8RGJ9yXlUrezuIm+V8ae3Si2v1gDPAYue2kSl9PhHJ6BwzKzGVcQwpMjulz2t8h90RmHTF+c3/HxFZ7fzck0idciKy3HkXsV5ESjrXP5Fg/ZciEpDIvpEiMthZb7mIlHCuLyqOeRwuzedQxLn+YRHZICLrRGSBc109EZnlvIPpBrzsPOe9ItJPRF4RkTIisvyq61rvXK4qIn+LyCoRmZPYyJIiMl5EPhaR+cBgEakmIovFMSb/YhEp7XwTdwDwiPP8j4hINnGMd7/CWTexEVuNv/H22Nv2sU9SHxxvxq51fqY712UFsjiXS+J4uxQgBOf47cAw4HHncibgNqAM8DMQ6Fw/AngykXNGAn2cy08Cs5zLPwNPOZc7AT85l/8FCjqXczj/rJdgv37AKwmO7yo7r6uYc/k14E0cb5AuBvI41z+C423aq+McD8wCApzl7EBG53Ij4Afnckfg8wT7DQSeuBQvsBXI5u2ftX28+7GmIZOWJdY0FAh8LiKVcSSKUonstwToIyKFgB9VdZuINASqAiucQ2zcBiQ15tK3Cf78xLlcE3jAuTwRxxwHAIuA8SLyHfDjzVwcjoHS2gGDcPzCfwQojWOwvLnOOAOApMaRmaaqcc7lYOBr592P4hySIBFNgNYi8oqznAUoAmy6ydiND7FEYNKbl4GDQCUcTZvXTDijqpNFZBlwHzBHRJ7BMXTv16r6uhvn0CSWr6mjqt1EpLrzXGudCcpdU3GM/fSj41C6TUQqAP+pak039j+bYPkdYL6q3u9skvoriX0EeFBVt9xEnMbHWR+BSW+Cgf3qGGu+A45vzFcQkWLADlUdimOkxorAPOAhEcnrrJNTkp63+ZEEfy5xLi/m8uCEjwMLnccprqrLVLUvcIQrhwcGOI1jSOxrqOp2HHc1b+FICuAYNjqPOMbVR0QCRaRcEnEmFAzsdS53vM755wA9xXm7IY5RaY2fs0Rg0psRwFMishRHs9DZROo8AmwQkbVAGI5p/TbiaIP/3dkpOxdIanq/zM47ihdx3IEAvAA87dy3g3MbwIci8q/z0dUFOObUTehn4P5LncWJnGsq8ASXx9O/iGNo9MEisg5HP8I1HeKJ+AB4X0QWcWVynA+UvdRZjOPOIRBY74z5HTeObXycjT5qTALimMQmXFWPeDsWY1KL3REYY4yfszsCY4zxc3ZHYIwxfs4SgTHG+DlLBMYY4+csERhjjJ+zRGCMMX7u/wE44O6gPEOI2QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Meta-label\n",
    "# Performance Metrics\n",
    "y_prob = rf.predict_proba(X_test)[:, 1] #here we are only interested in True positive\n",
    "y_pred = rf.predict(X_test)\n",
    "\n",
    "p('Matrix test report for primary model & secondary model\\n')\n",
    "\n",
    "rs.report_matrix(actual_data = y_test, \n",
    "                 prediction_data = y_pred, \n",
    "                 ROC = y_prob)\n",
    "\n",
    "rs.feat_imp(rf, X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "**Now we start to create only primary model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Wei_X\\Desktop\\Python\\research\\Labels\\triple_barrier_method.py:75: UserWarning: Data and events index shape must be same, reindex data to fit events\n",
      "  warnings.warn('Data and events index shape must be same, reindex data to fit events')\n",
      "C:\\Users\\Wei_X\\Desktop\\Python\\research\\Labels\\triple_barrier_method.py:112: UserWarning: Not Recommended: No side prediction provided\n",
      "  warnings.warn('Not Recommended: No side prediction provided')\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[                                             t1  sl  pt\n",
      "2015-09-09 13:36:53.472 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:40:56.133 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:45:08.189 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:51:11.450 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 14:02:04.210 2015-09-10 14:05:19.003 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-04-05 01:16:26.261 2016-04-06 02:45:07.823 NaT NaT\n",
      "2016-04-05 06:02:04.305 2016-04-06 07:07:24.337 NaT NaT\n",
      "2016-04-05 07:11:07.085 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 07:37:56.273 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 08:24:13.547 2016-04-06 09:05:34.702 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns]] this out\n",
      "[                                             t1  sl  pt\n",
      "2015-09-09 13:36:53.472 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:40:56.133 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:45:08.189 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:51:11.450 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 14:02:04.210 2015-09-10 14:05:19.003 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-04-05 01:16:26.261 2016-04-06 02:45:07.823 NaT NaT\n",
      "2016-04-05 06:02:04.305 2016-04-06 07:07:24.337 NaT NaT\n",
      "2016-04-05 07:11:07.085 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 07:37:56.273 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 08:24:13.547 2016-04-06 09:05:34.702 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns],                                              t1  sl  pt\n",
      "2015-01-05 14:54:26.286 2015-01-06 15:16:04.445 NaT NaT\n",
      "2015-01-05 14:57:13.616 2015-01-06 15:16:04.445 NaT NaT\n",
      "2015-01-05 15:01:57.494 2015-01-06 15:16:04.445 NaT NaT\n",
      "2015-01-05 15:07:29.012 2015-01-06 15:16:04.445 NaT NaT\n",
      "2015-01-05 15:13:09.655 2015-01-06 15:16:04.445 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2015-09-09 08:33:28.634 2015-09-10 09:33:29.450 NaT NaT\n",
      "2015-09-09 10:21:19.955 2015-09-10 11:43:14.539 NaT NaT\n",
      "2015-09-09 12:02:05.852 2015-09-10 12:19:25.229 NaT NaT\n",
      "2015-09-09 13:06:01.330 2015-09-10 13:30:20.359 NaT NaT\n",
      "2015-09-09 13:31:24.524 2015-09-10 13:35:08.110 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns]] this out\n",
      "[                                             t1  sl  pt\n",
      "2015-09-09 13:36:53.472 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:40:56.133 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:45:08.189 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 13:51:11.450 2015-09-10 13:59:08.553 NaT NaT\n",
      "2015-09-09 14:02:04.210 2015-09-10 14:05:19.003 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-04-05 01:16:26.261 2016-04-06 02:45:07.823 NaT NaT\n",
      "2016-04-05 06:02:04.305 2016-04-06 07:07:24.337 NaT NaT\n",
      "2016-04-05 07:11:07.085 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 07:37:56.273 2016-04-06 08:05:23.022 NaT NaT\n",
      "2016-04-05 08:24:13.547 2016-04-06 09:05:34.702 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns],                                              t1  sl  pt\n",
      "2015-01-05 14:54:26.286 2015-01-06 15:16:04.445 NaT NaT\n",
      "2015-01-05 14:57:13.616 2015-01-06 15:16:04.445 NaT NaT\n",
      "2015-01-05 15:01:57.494 2015-01-06 15:16:04.445 NaT NaT\n",
      "2015-01-05 15:07:29.012 2015-01-06 15:16:04.445 NaT NaT\n",
      "2015-01-05 15:13:09.655 2015-01-06 15:16:04.445 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2015-09-09 08:33:28.634 2015-09-10 09:33:29.450 NaT NaT\n",
      "2015-09-09 10:21:19.955 2015-09-10 11:43:14.539 NaT NaT\n",
      "2015-09-09 12:02:05.852 2015-09-10 12:19:25.229 NaT NaT\n",
      "2015-09-09 13:06:01.330 2015-09-10 13:30:20.359 NaT NaT\n",
      "2015-09-09 13:31:24.524 2015-09-10 13:35:08.110 NaT NaT\n",
      "\n",
      "[5142 rows x 3 columns],                                              t1  sl  pt\n",
      "2016-04-05 09:17:57.890 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 10:11:56.123 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 11:26:40.687 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 12:10:26.367 2016-04-06 13:14:38.661 NaT NaT\n",
      "2016-04-05 12:55:36.469 2016-04-06 13:14:38.661 NaT NaT\n",
      "...                                         ...  ..  ..\n",
      "2016-12-30 20:22:33.456                     NaT NaT NaT\n",
      "2016-12-30 20:32:57.188                     NaT NaT NaT\n",
      "2016-12-30 20:44:21.481                     NaT NaT NaT\n",
      "2016-12-30 20:50:57.567                     NaT NaT NaT\n",
      "2016-12-30 20:55:33.160                     NaT NaT NaT\n",
      "\n",
      "[5141 rows x 3 columns]]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-05-24 19:48:27.258258 33.33% _pt_sl_t1 done after 0.41 minutes. Remaining 0.83 minutes.\r",
      "2020-05-24 19:48:27.258258 66.67% _pt_sl_t1 done after 0.41 minutes. Remaining 0.21 minutes.\r"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " this out\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-05-24 19:48:27.445715 100.0% _pt_sl_t1 done after 0.42 minutes. Remaining 0.0 minutes.\n"
     ]
    }
   ],
   "source": [
    "events1 = rs.cs_filter(pri_dollar['close'], \n",
    "                    limit = pri_dollar['volatility'].mean())\n",
    "\n",
    "vb1 = rs.vert_barrier(data = pri_dollar['close'], \n",
    "                 events = events1, \n",
    "                 period = 'days', \n",
    "                 freq = 1)\n",
    "\n",
    "tb1 = rs.tri_barrier(data = pri_dollar['close'], \n",
    "                events = events1, \n",
    "                trgt = pri_dollar['volatility'], \n",
    "                min_req = 0.002, \n",
    "                num_threads = 3, \n",
    "                ptSl = [0,2], #change ptSl into [0,2]\n",
    "                t1 = vb1, \n",
    "                side = None)\n",
    "\n",
    "m_label1 = rs.meta_label(data = pri_dollar['close'],\n",
    "                      events = tb1,\n",
    "                      drop = 0.05) # take note we do not have a side hence we need to drop something"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='entropy',\n",
       "                       max_depth=7, max_features='auto', max_leaf_nodes=None,\n",
       "                       min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                       min_samples_leaf=1, min_samples_split=2,\n",
       "                       min_weight_fraction_leaf=0.0, n_estimators=500,\n",
       "                       n_jobs=None, oob_score=False, random_state=42, verbose=0,\n",
       "                       warm_start=False)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Random Forest Model\n",
    "rf = RandomForestClassifier(max_depth=max_depth, \n",
    "                            n_estimators=n_estimators,\n",
    "                            criterion='entropy', \n",
    "                            class_weight = None, #This will be cover in next few chapters\n",
    "                            random_state=c_random_state)\n",
    "\n",
    "X = pri_dollar.reindex(m_label1.index) # this dataframe only contain all our features\n",
    "y = m_label1['bin']\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, shuffle=False)\n",
    "\n",
    "rf.fit(X_train, y_train.values.ravel())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Matrix training report for primary model only\n",
      "\n",
      " Classification Report\n",
      "=======================================================\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        -1.0       0.71      0.43      0.54      4965\n",
      "         1.0       0.63      0.85      0.73      5748\n",
      "\n",
      "    accuracy                           0.66     10713\n",
      "   macro avg       0.67      0.64      0.63     10713\n",
      "weighted avg       0.67      0.66      0.64     10713\n",
      "\n",
      "\n",
      " Confusion Matrix\n",
      "=======================================================\n",
      "[[TN, FP]\n",
      " [FN, TP]]\n",
      "\n",
      "[[2133 2832]\n",
      " [ 851 4897]]\n",
      "\n",
      " Accuracy Score\n",
      "=======================================================\n",
      "0.6562120787827873\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Performance Metrics\n",
    "y_prob = rf.predict_proba(X_train)[:, 1] #here we are only interested in True positive\n",
    "y_pred = rf.predict(X_train)\n",
    "\n",
    "p('Matrix training report for primary model only\\n')\n",
    "\n",
    "rs.report_matrix(actual_data = y_train, # we need to use our train data from train_test_split\n",
    "                 prediction_data = y_pred, \n",
    "                 ROC = y_prob)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Matrix test report for primary model only\n",
      "\n",
      " Classification Report\n",
      "=======================================================\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        -1.0       0.46      0.73      0.56      2027\n",
      "         1.0       0.60      0.31      0.41      2565\n",
      "\n",
      "    accuracy                           0.50      4592\n",
      "   macro avg       0.53      0.52      0.49      4592\n",
      "weighted avg       0.54      0.50      0.48      4592\n",
      "\n",
      "\n",
      " Confusion Matrix\n",
      "=======================================================\n",
      "[[TN, FP]\n",
      " [FN, TP]]\n",
      "\n",
      "[[1484  543]\n",
      " [1762  803]]\n",
      "\n",
      " Accuracy Score\n",
      "=======================================================\n",
      "0.4980400696864111\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Meta-label\n",
    "# Performance Metrics\n",
    "y_prob = rf.predict_proba(X_test)[:, 1] #here we are only interested in True positive\n",
    "y_pred = rf.predict(X_test)\n",
    "\n",
    "p('Matrix test report for primary model only\\n')\n",
    "\n",
    "rs.report_matrix(actual_data = y_test, \n",
    "                 prediction_data = y_pred, \n",
    "                 ROC = y_prob)\n",
    "\n",
    "rs.feat_imp(rf, X)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Based on our matrix report\n",
    "\n",
    "The comparisons was made in ceteris paribus condition as much as possible.\n",
    "\n",
    "All comparison will be made based on test data only, train data will be excluded.\n",
    "    \n",
    "### Accuracy comparison\n",
    "\n",
    "**Accuracy is sum of True Positive and True Negative divided by overall set of items**\n",
    "\n",
    "There is an improvement in accuracy rate when meta-label (Primary & secondary model score yield 0.53), with original yielding only 0.48.\n",
    "\n",
    "Accuracy +0.05 improvement (10% increase) from the original, +0.0342 improvement (6.8% increase) from primary model only.\n",
    "\n",
    "However, in order to correctly use this method. Dr Marco Lopez De Prado did mentioned the below:\n",
    "\n",
    ">\"First, we build a model that achieve high recall, even if precision is not particularly high.\n",
    ">\n",
    "> Second correct for low precision by applying meta-label to the positives predicted by the primary model.\" \n",
    ">\n",
    "> Advances in Financial Machine Learning, page 52\n",
    "\n",
    "However, for primary model only case. (-1,1) are consider price actions label, which in my opinion does not seem to work well with ML. But it does improve accuracy score by a small margin against original data with no labels. \n",
    "\n",
    "In our case, we filtered out labels that touched vertical barrier first from primary model only.\n",
    "\n",
    "### F1 scores comparison\n",
    "**Measures the efficiency of classifer (Harmonic mean of both precision and recall)**\n",
    "\n",
    "Using both primary and secondary models to identify True Positive yields a score of 0.6 (for both long and short).\n",
    "\n",
    "while using primary model only which gives 0.56 (for short) and 0.4 (for long).\n",
    "\n",
    "F1 score +0.04 (7% increase/ short) and +0.2 (50% increase/ long) improvement, when compared against primary model only.\n",
    "\n",
    ">\"Meta-labeling is particularly helpful when you want to achieve higher F1-scores.\"\n",
    ">\n",
    "> Advances in Financial Machine Learning, page 52\n",
    "\n",
    "### Other observations\n",
    "\n",
    "<ins>Stationarity absolute return series</ins> as a optional key feature, does not seem relevant at all since the ML does not recognize it after log price absolute change. Hence it is lowly ranked in feature importance graph.\n",
    "\n",
    "**For secondary model**\n",
    "\n",
    "Crossing averages and volatility does seem to be at the top of the key features importance. This seems to say, ML model recognize that as key feature for predictions, while auto-correlation does not seem that important.\n",
    "\n",
    "The ML model did realised, we were using those indicators as our primary models.\n",
    "\n",
    "**For Primary model**\n",
    "\n",
    "The ML model only recognize we were using volatility as our benchmark (tri_barrier func cs_filter as trgt), when we did not use any (0,1) meta-labels.\n",
    "\n",
    "### Conclusion\n",
    "\n",
    "To get a higher F1 score and better accuracy, quants should use both primary (To let it decide bet direction) while using a secondary model to decide bet size (To bet or not).\n",
    "\n",
    "Stationarity is an important concept, especially to mean-reversion strategy. As Stationary series act as an anchor which the strategy will revert to eventually.\n",
    "\n",
    "We will cover more with other examples regarding stationarity in the next chapter."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}