train_model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Some markdown is never a bad idea\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import time\n",
    "# visualization modules\n",
    "import matplotlib as mpl\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "# Machine learning models\n",
    "from sklearn.model_selection import train_test_split\n",
    "import xgboost as xgb\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.svm import SVC\n",
    "import pickle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "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>Unnamed: 0</th>\n",
       "      <th>Date</th>\n",
       "      <th>HomeTeam</th>\n",
       "      <th>AwayTeam</th>\n",
       "      <th>FTHG</th>\n",
       "      <th>FTAG</th>\n",
       "      <th>FTR</th>\n",
       "      <th>HTHG</th>\n",
       "      <th>HTAG</th>\n",
       "      <th>HTR</th>\n",
       "      <th>...</th>\n",
       "      <th>HY</th>\n",
       "      <th>AY</th>\n",
       "      <th>HR</th>\n",
       "      <th>AR</th>\n",
       "      <th>B365H</th>\n",
       "      <th>B365D</th>\n",
       "      <th>B365A</th>\n",
       "      <th>BbAvH</th>\n",
       "      <th>BbAvD</th>\n",
       "      <th>BbAvA</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>2009-08-15</td>\n",
       "      <td>Aston Villa</td>\n",
       "      <td>Wigan</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>-1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>A</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1.67</td>\n",
       "      <td>3.60</td>\n",
       "      <td>5.50</td>\n",
       "      <td>1.66</td>\n",
       "      <td>3.51</td>\n",
       "      <td>5.33</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2009-08-15</td>\n",
       "      <td>Blackburn</td>\n",
       "      <td>Man City</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>-1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>A</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3.60</td>\n",
       "      <td>3.25</td>\n",
       "      <td>2.10</td>\n",
       "      <td>3.37</td>\n",
       "      <td>3.24</td>\n",
       "      <td>2.12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>2009-08-15</td>\n",
       "      <td>Bolton</td>\n",
       "      <td>Sunderland</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>A</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2.25</td>\n",
       "      <td>3.25</td>\n",
       "      <td>3.25</td>\n",
       "      <td>2.24</td>\n",
       "      <td>3.20</td>\n",
       "      <td>3.15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>2009-08-15</td>\n",
       "      <td>Chelsea</td>\n",
       "      <td>Hull</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>D</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1.17</td>\n",
       "      <td>6.50</td>\n",
       "      <td>21.00</td>\n",
       "      <td>1.17</td>\n",
       "      <td>6.26</td>\n",
       "      <td>16.39</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>2009-08-15</td>\n",
       "      <td>Everton</td>\n",
       "      <td>Arsenal</td>\n",
       "      <td>1</td>\n",
       "      <td>6</td>\n",
       "      <td>-1</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>A</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3.20</td>\n",
       "      <td>3.25</td>\n",
       "      <td>2.30</td>\n",
       "      <td>3.07</td>\n",
       "      <td>3.21</td>\n",
       "      <td>2.28</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",
       "      <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>3795</th>\n",
       "      <td>3795</td>\n",
       "      <td>2019-12-05</td>\n",
       "      <td>Liverpool</td>\n",
       "      <td>Wolves</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>H</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1.30</td>\n",
       "      <td>6.00</td>\n",
       "      <td>11.00</td>\n",
       "      <td>1.30</td>\n",
       "      <td>5.62</td>\n",
       "      <td>10.17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3796</th>\n",
       "      <td>3796</td>\n",
       "      <td>2019-12-05</td>\n",
       "      <td>Man United</td>\n",
       "      <td>Cardiff</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>-1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>A</td>\n",
       "      <td>...</td>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1.28</td>\n",
       "      <td>6.50</td>\n",
       "      <td>11.00</td>\n",
       "      <td>1.28</td>\n",
       "      <td>6.18</td>\n",
       "      <td>10.10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3797</th>\n",
       "      <td>3797</td>\n",
       "      <td>2019-12-05</td>\n",
       "      <td>Southampton</td>\n",
       "      <td>Huddersfield</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>H</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1.44</td>\n",
       "      <td>4.75</td>\n",
       "      <td>8.50</td>\n",
       "      <td>1.42</td>\n",
       "      <td>4.81</td>\n",
       "      <td>7.64</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3798</th>\n",
       "      <td>3798</td>\n",
       "      <td>2019-12-05</td>\n",
       "      <td>Tottenham</td>\n",
       "      <td>Everton</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>H</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2.20</td>\n",
       "      <td>3.50</td>\n",
       "      <td>3.50</td>\n",
       "      <td>2.09</td>\n",
       "      <td>3.51</td>\n",
       "      <td>3.58</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3799</th>\n",
       "      <td>3799</td>\n",
       "      <td>2019-12-05</td>\n",
       "      <td>Watford</td>\n",
       "      <td>West Ham</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>-1</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>A</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>2.25</td>\n",
       "      <td>3.75</td>\n",
       "      <td>3.20</td>\n",
       "      <td>2.20</td>\n",
       "      <td>3.73</td>\n",
       "      <td>3.13</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3800 rows × 28 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      Unnamed: 0        Date     HomeTeam      AwayTeam  FTHG  FTAG  FTR  \\\n",
       "0              0  2009-08-15  Aston Villa         Wigan     0     2   -1   \n",
       "1              1  2009-08-15    Blackburn      Man City     0     2   -1   \n",
       "2              2  2009-08-15       Bolton    Sunderland     0     1   -1   \n",
       "3              3  2009-08-15      Chelsea          Hull     2     1    1   \n",
       "4              4  2009-08-15      Everton       Arsenal     1     6   -1   \n",
       "...          ...         ...          ...           ...   ...   ...  ...   \n",
       "3795        3795  2019-12-05    Liverpool        Wolves     2     0    1   \n",
       "3796        3796  2019-12-05   Man United       Cardiff     0     2   -1   \n",
       "3797        3797  2019-12-05  Southampton  Huddersfield     1     1    0   \n",
       "3798        3798  2019-12-05    Tottenham       Everton     2     2    0   \n",
       "3799        3799  2019-12-05      Watford      West Ham     1     4   -1   \n",
       "\n",
       "      HTHG  HTAG HTR  ...  HY  AY  HR  AR  B365H  B365D  B365A  BbAvH  BbAvD  \\\n",
       "0        0     1   A  ...   2   2   0   0   1.67   3.60   5.50   1.66   3.51   \n",
       "1        0     1   A  ...   2   1   0   0   3.60   3.25   2.10   3.37   3.24   \n",
       "2        0     1   A  ...   2   1   0   0   2.25   3.25   3.25   2.24   3.20   \n",
       "3        1     1   D  ...   1   2   0   0   1.17   6.50  21.00   1.17   6.26   \n",
       "4        0     3   A  ...   0   0   0   0   3.20   3.25   2.30   3.07   3.21   \n",
       "...    ...   ...  ..  ...  ..  ..  ..  ..    ...    ...    ...    ...    ...   \n",
       "3795     1     0   H  ...   0   2   0   0   1.30   6.00  11.00   1.30   5.62   \n",
       "3796     0     1   A  ...   3   3   0   0   1.28   6.50  11.00   1.28   6.18   \n",
       "3797     1     0   H  ...   0   1   0   0   1.44   4.75   8.50   1.42   4.81   \n",
       "3798     1     0   H  ...   0   2   0   0   2.20   3.50   3.50   2.09   3.51   \n",
       "3799     0     2   A  ...   1   0   1   0   2.25   3.75   3.20   2.20   3.73   \n",
       "\n",
       "      BbAvA  \n",
       "0      5.33  \n",
       "1      2.12  \n",
       "2      3.15  \n",
       "3     16.39  \n",
       "4      2.28  \n",
       "...     ...  \n",
       "3795  10.17  \n",
       "3796  10.10  \n",
       "3797   7.64  \n",
       "3798   3.58  \n",
       "3799   3.13  \n",
       "\n",
       "[3800 rows x 28 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Reading the file from disk\n",
    "matches = pd.read_csv('data/seasons_merged.csv')\n",
    "letter_to_result = {'H': 1, 'D': 0, 'A': -1}\n",
    "\n",
    "matches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "def get_n_last_matches(matches, date, team, n = 10):\n",
    "    '''\n",
    "    Get the last n matches of a given team.\n",
    "    '''\n",
    "    # All matches with a given team\n",
    "    team_matches = matches[(matches['HomeTeam'] == team) | (matches['AwayTeam'] == team)]\n",
    "    \n",
    "    #Filter n last matches from team matches\n",
    "    n_last_matches = (team_matches[team_matches.Date < date]\n",
    "                      .sort_values(by = 'Date', ascending = False)\n",
    "                      .iloc[0:n,:])\n",
    "    \n",
    "    return n_last_matches\n",
    "\n",
    "\n",
    "def get_n_last_matches_against_each_other(matches, date, home_team, away_team, n = 5):\n",
    "    ''' \n",
    "    Get the last n matches of two given teams. If possible, else\n",
    "    get all matches available\n",
    "    '''\n",
    "    \n",
    "    home_matches = matches[(matches['HomeTeam'] == home_team) & (matches['AwayTeam'] == away_team)]    \n",
    "    away_matches = matches[(matches['HomeTeam'] == away_team) & (matches['AwayTeam'] == home_team)]  \n",
    "    total_matches = pd.concat([home_matches, away_matches])\n",
    "    \n",
    "    # Get last n matches, if possible:\n",
    "    try:    \n",
    "        last_matches = (total_matches[total_matches.Date < date]\n",
    "                        .sort_values(by = 'Date', ascending = False)\n",
    "                        .iloc[0:n,:])\n",
    "    except:  # If there are less than n matches\n",
    "        last_matches = (total_matches[total_matches.Date < date]\n",
    "                        .sort_values(by = 'Date', ascending = False)\n",
    "                        .iloc[0:total_matches.shape[0],:])\n",
    "\n",
    "    return last_matches\n",
    "\n",
    "\n",
    "def get_goals(matches, team):\n",
    "    '''\n",
    "    Get total number of goals,a specfic team has scored \n",
    "    from a dataframe of specific matches\n",
    "    '''\n",
    "    home_goals = matches.FTHG[matches.HomeTeam == team].sum()\n",
    "    away_goals = matches.FTAG[matches.AwayTeam == team].sum()\n",
    "\n",
    "    return home_goals + away_goals\n",
    "\n",
    "def get_concealed_goals(matches, team):\n",
    "    '''\n",
    "    Get all the goals, concealed of a specfic team from a dataframe of specific matches\n",
    "    '''\n",
    "    home_goals = matches.FTAG[matches.HomeTeam == team].sum()\n",
    "    away_goals = matches.FTHG[matches.AwayTeam == team].sum()\n",
    "\n",
    "    return home_goals + away_goals\n",
    "\n",
    "def get_wins(matches, team):\n",
    "    '''\n",
    "    Get the number of wins of a specfic team from a dataframe of specific matches.\n",
    "    '''\n",
    "    home_wins = matches[(matches.FTR == 1) & (matches.HomeTeam == team)].shape[0]\n",
    "    away_wins = matches[(matches.FTR == -1) & (matches.AwayTeam == team)].shape[0]\n",
    "\n",
    "    return home_wins + away_wins\n",
    "\n",
    "\n",
    "def coefficients_to_probability(matches):\n",
    "    '''\n",
    "    Converts betting platform coefficient(1 < x) with % of income\n",
    "    into a probability coefficient(0 < x < 1)\n",
    "    '''\n",
    "    # How many profit betting companies make on bets\n",
    "    matches['profit_B365'] = sum((1/matches['B365H'], 1/matches['B365D'], 1/matches['B365A']))\n",
    "    matches['profit_BbAv'] = sum((1/matches['BbAvA'], 1/matches['BbAvD'], 1/matches['BbAvH']))\n",
    "\n",
    "    # Converting all betting coefficients into probabilities of homw/draw/away:\n",
    "    for betting_column in ['B365H', 'B365D', 'B365A', 'BbAvH', 'BbAvD', 'BbAvA']:\n",
    "        matches[betting_column] = 1 / (matches[betting_column] * matches['profit_' + betting_column[:-1]])\n",
    "\n",
    "    return matches\n",
    "# get_n_last_matches_against_each_other(matches, '2020-12-15', 'Liverpool', 'Tottenham', 10)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# Create features, based on which, the model would train and predict results\n",
    "\n",
    "def get_features_for_match(match, matches, n1=10, n2=3):\n",
    "    '''\n",
    "    Creates a special set of features for each match, if possible(10 last matches\n",
    "    and 3 last matches against each other)\n",
    "    '''\n",
    "    match_date = match.Date\n",
    "    home_team = match.HomeTeam\n",
    "    away_team = match.AwayTeam\n",
    "    # Get n1 last matches of 2 teams\n",
    "    home_last = get_n_last_matches(matches, match_date, home_team, n=n1)\n",
    "    away_last = get_n_last_matches(matches, match_date, away_team, n=n1)\n",
    "    # Get last n2 matches against each other\n",
    "    home_last_against = get_n_last_matches_against_each_other(matches, match_date, home_team, away_team, n=n2)\n",
    "    away_last_against = get_n_last_matches_against_each_other(matches, match_date, away_team, home_team, n=n2)\n",
    "    # Goals stuff\n",
    "    home_goals = get_goals(home_last, home_team)\n",
    "    away_goals = get_goals(away_last, away_team)\n",
    "    home_goals_conceided = get_concealed_goals(home_last, home_team)\n",
    "    away_goals_conceided = get_concealed_goals(away_last, away_team)\n",
    "\n",
    "    res = pd.DataFrame()\n",
    "    res.loc[0, 'H_goal_diff'] = home_goals - home_goals_conceided\n",
    "    res.loc[0, 'A_goal_diff'] = away_goals - away_goals_conceided\n",
    "    res.loc[0, 'H_win'] = get_wins(home_last, home_team) \n",
    "    res.loc[0, 'A_win'] = get_wins(away_last, away_team)\n",
    "    res.loc[0, 'H_win_against'] = get_wins(home_last_against, home_team)\n",
    "    res.loc[0, 'A_win_against'] = get_wins(away_last_against, away_team)\n",
    "    # TODO ПОПРООБУВАТИ ЩЕ ЯКІСЬ КРИТЕРІЇ ПОТЕСТУВАТИ\n",
    "#     print(result.loc[0])\n",
    "    return res.loc[0]\n",
    "\n",
    "\n",
    "def create_features(matches):\n",
    "    '''\n",
    "    Iterate throu all matches, create features for every single of them\n",
    "    if possible and aggregate them together\n",
    "    '''\n",
    "    print('Generating features... Please wait for one or two minutes')\n",
    "    # Creates dataframe with features for all matches\n",
    "    matches_features = matches.apply(lambda x: get_features_for_match(x, matches, n1=20, n2=3), axis=1)\n",
    "    return matches_features\n",
    "\n",
    "def explore_features(features, inputs, path):\n",
    "    '''\n",
    "    Explore data by plotting KDE graphs.\n",
    "    '''\n",
    "    fig = plt.figure(1)\n",
    "    fig.subplots_adjust(bottom= -1, left=0.025, top = 2, right=0.975)\n",
    "    #Loop through features    \n",
    "    i = 1\n",
    "    for col in features.columns:\n",
    "        sns.set_style(\"whitegrid\")\n",
    "        sns.set_context(\"paper\", font_scale = 0.5, rc={\"lines.linewidth\": 1})\n",
    "        plt.subplot(5,5,0 + i)\n",
    "        j = i - 1\n",
    "        #Plot KDE for all labels\n",
    "        try:\n",
    "            sns.distplot(inputs[inputs['label'] == 'H'].iloc[:,j], hist = False, label = 'Home')\n",
    "            sns.distplot(inputs[inputs['label'] == 'D'].iloc[:,j], hist = False, label = 'Draw')\n",
    "            sns.distplot(inputs[inputs['label'] == 'A'].iloc[:,j], hist = False, label = 'Away')\n",
    "            plt.legend();\n",
    "        except ValueError as e:\n",
    "            print(e)\n",
    "        i = i + 1\n",
    "    #Define plot format    \n",
    "    DefaultSize = fig.get_size_inches()\n",
    "    fig.set_size_inches((DefaultSize[0]*1.2, DefaultSize[1]*1.2))\n",
    "    plt.show()\n",
    "    #Compute and print label weights\n",
    "    labels = inputs.loc[:,'label']\n",
    "    class_weights = labels.value_counts() / len(labels)\n",
    "    print(class_weights)\n",
    "    #Store description of all features\n",
    "    feature_details = features.describe().transpose()\n",
    "    #Return feature details\n",
    "    return feature_details"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# Getting features ready\n",
    "\n",
    "# If you want to create features instead of reading them from file, uncomment:\n",
    "\n",
    "# time_start = time.time()\n",
    "# features = create_features(matches)\n",
    "# print('Generated features in', time.time() - time_start, 'sec.')\n",
    "# features.iloc[100:].to_csv('data/features.csv')\n",
    "\n",
    "# Reading files from the file\n",
    "features = pd.read_csv('data/features.csv')\n",
    "# features = features.drop('Unnamed: 0', axis=1)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Printing statistics to see the correlations:\n",
    "```python3\n",
    "statistic = explore_features(features, inputs, '/soccer/')\n",
    "statistic\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 518.4x345.6 with 7 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "D    0.413889\n",
      "H    0.346944\n",
      "A    0.239167\n",
      "Name: label, dtype: float64\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>count</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>25%</th>\n",
       "      <th>50%</th>\n",
       "      <th>75%</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <td>3700.0</td>\n",
       "      <td>1949.500000</td>\n",
       "      <td>1068.242326</td>\n",
       "      <td>100.0</td>\n",
       "      <td>1024.75</td>\n",
       "      <td>1949.5</td>\n",
       "      <td>2874.25</td>\n",
       "      <td>3799.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H_goal_diff</th>\n",
       "      <td>3700.0</td>\n",
       "      <td>0.236757</td>\n",
       "      <td>14.861263</td>\n",
       "      <td>-37.0</td>\n",
       "      <td>-11.00</td>\n",
       "      <td>-2.0</td>\n",
       "      <td>10.00</td>\n",
       "      <td>53.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A_goal_diff</th>\n",
       "      <td>3700.0</td>\n",
       "      <td>0.559459</td>\n",
       "      <td>14.841652</td>\n",
       "      <td>-37.0</td>\n",
       "      <td>-10.00</td>\n",
       "      <td>-2.0</td>\n",
       "      <td>10.00</td>\n",
       "      <td>52.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H_win</th>\n",
       "      <td>3700.0</td>\n",
       "      <td>7.309730</td>\n",
       "      <td>3.570593</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5.00</td>\n",
       "      <td>7.0</td>\n",
       "      <td>10.00</td>\n",
       "      <td>19.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A_win</th>\n",
       "      <td>3700.0</td>\n",
       "      <td>7.385405</td>\n",
       "      <td>3.561851</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5.00</td>\n",
       "      <td>7.0</td>\n",
       "      <td>10.00</td>\n",
       "      <td>19.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H_win_against</th>\n",
       "      <td>3700.0</td>\n",
       "      <td>0.808108</td>\n",
       "      <td>0.888142</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.00</td>\n",
       "      <td>3.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A_win_against</th>\n",
       "      <td>3700.0</td>\n",
       "      <td>0.886486</td>\n",
       "      <td>0.906604</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.00</td>\n",
       "      <td>3.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                count         mean          std    min      25%     50%  \\\n",
       "Unnamed: 0     3700.0  1949.500000  1068.242326  100.0  1024.75  1949.5   \n",
       "H_goal_diff    3700.0     0.236757    14.861263  -37.0   -11.00    -2.0   \n",
       "A_goal_diff    3700.0     0.559459    14.841652  -37.0   -10.00    -2.0   \n",
       "H_win          3700.0     7.309730     3.570593    0.0     5.00     7.0   \n",
       "A_win          3700.0     7.385405     3.561851    0.0     5.00     7.0   \n",
       "H_win_against  3700.0     0.808108     0.888142    0.0     0.00     1.0   \n",
       "A_win_against  3700.0     0.886486     0.906604    0.0     0.00     1.0   \n",
       "\n",
       "                   75%     max  \n",
       "Unnamed: 0     2874.25  3799.0  \n",
       "H_goal_diff      10.00    53.0  \n",
       "A_goal_diff      10.00    52.0  \n",
       "H_win            10.00    19.0  \n",
       "A_win            10.00    19.0  \n",
       "H_win_against     1.00     3.0  \n",
       "A_win_against     1.00     3.0  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Determining labels, which we want to train on, and then predict them\n",
    "\n",
    "labels = matches.loc[:, 'HTR']\n",
    "labels.name = 'label'\n",
    "labels = labels.iloc[100:]\n",
    "# coefficients = coefficients_to_probability(matches)[['BbAvH','BbAvD','BbAvA']]\n",
    "# features = pd.merge(features, coefficients, left_index=True, right_index=True)\n",
    "# statistic = explore_features(features, features.deo, '/soccer/')\n",
    "# statistic\n",
    "\n",
    "inputs = pd.merge(features, labels, left_index=True, right_index=True)\n",
    "statistic = explore_features(features, inputs, '/soccer/')\n",
    "statistic"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Machine learning part \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# Splitting the data into training and testing parts\n",
    "# Standardising the data.\n",
    "from sklearn.preprocessing import scale\n",
    "\n",
    "#Center to the mean and component wise scale to unit variance.\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(features, labels, \n",
    "                                                    test_size = 400,\n",
    "                                                    random_state = 2,\n",
    "                                                    stratify = labels)\n",
    "\n",
    "# Classifiers\n",
    "# Defyning classifiers:\n",
    "\n",
    "clf_A = LogisticRegression(random_state = 42, max_iter=1200000)\n",
    "clf_B = SVC(random_state = 912, kernel='rbf', max_iter=1200000)\n",
    "clf_C = xgb.XGBClassifier(seed = 82)\n",
    "\n",
    "# After maxyzizing, the following parameters proved to behave the best:\n",
    "clf_boosted = xgb.XGBClassifier(base_score=0.5, colsample_bylevel=1, colsample_bytree=0.8,\n",
    "       gamma=0.4, learning_rate=0.1, max_delta_step=0, max_depth=3,\n",
    "       min_child_weight=3, missing=None, n_estimators=40, nthread=-1,\n",
    "       objective='binary:logistic', reg_alpha=1e-05, reg_lambda=1,\n",
    "       scale_pos_weight=1, seed=2, silent=True, subsample=0.8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import f1_score\n",
    "\n",
    "def train_classifier(clf, X_train, y_train):\n",
    "    ''' \n",
    "    Fits and overall trains a classifier to the training data. \n",
    "    '''\n",
    "    \n",
    "    # Start the clock, train the classifier, then stop the clock\n",
    "    start = time.time()\n",
    "    clf.fit(X_train, y_train)\n",
    "    end = time.time()\n",
    "    \n",
    "    # Print the results\n",
    "    print(f\"Trained model in {end - start} seconds\")\n",
    "\n",
    "    \n",
    "def predict_labels(clf, features, target):\n",
    "    '''\n",
    "    Makes predictions using a fit classifier based on F1 score. \n",
    "    '''\n",
    "    \n",
    "    # Start the clock, make predictions, then stop the clock\n",
    "    start = time.time()\n",
    "    y_pred = clf.predict(features)\n",
    "    \n",
    "    end = time.time()\n",
    "    # Print and return results\n",
    "    print(f\"Predicted in {end - start} seconds\")\n",
    "    \n",
    "    return f1_score(target, y_pred, pos_label='H'), sum(target == y_pred) / float(len(y_pred))\n",
    "\n",
    "\n",
    "def train_predict(clf, X_train, y_train, X_test, y_test):\n",
    "    ''' Train and predict using a classifer based on F1 score. '''\n",
    "    \n",
    "    # Indicate the classifier and the training set size\n",
    "    print(\"Training a {} using a training set size of {}. . .\".format(clf.__class__.__name__, len(X_train)))\n",
    "    \n",
    "    # Train the classifier\n",
    "    train_classifier(clf, X_train, y_train)\n",
    "    \n",
    "    # Print the results of prediction for both training and testing\n",
    "    f1, acc = predi\n",
    "\n",
    "\n",
    "    \n",
    "def predict_labels(clf, features, target):\n",
    "    '''\n",
    "    Makes predictions using a fit classifier based on F1 score. \n",
    "    '''\n",
    "    \n",
    "    # Start the clock, make predictions, then stop the clock\n",
    "    start = time.time()\n",
    "    y_pred = clf.predict(features)\n",
    "    \n",
    "    end = time.time()\n",
    "    # Print and return results\n",
    "    print(f\"Predicted in {end - start} seconds\")\n",
    "#     print(f1_score(target, y_pred, pos_label=1))\n",
    "    return f1_score(target, y_pred, pos_label=1), sum(target == y_pred) / float(len(y_pred))\n",
    "\n",
    "\n",
    "def train_predict(clf, X_train, y_train, X_test, y_test):\n",
    "    '''\n",
    "    Train and predict using a given classifer\n",
    "    '''\n",
    "    \n",
    "    # Indicate the classifier and the training set size\n",
    "    print(f\"Training a {clf.__class__.__name__} using a training set size of {len(X_train)}\")\n",
    "    \n",
    " # Train the classifier\n",
    "    train_classifier(clf, X_train, y_train)\n",
    "    \n",
    "    # Print the results of prediction for both training and testing\n",
    "    f1, acc = predict_labels(clf, X_train, y_train)\n",
    "    print(f\"Accyracy on training set: {acc}.\")\n",
    "    \n",
    "    f1, acc = predict_labels(clf, X_test, y_test)\n",
    "    print(f\"Accyracy on test set: {acc}.\")\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2447    0\n",
       "2316    0\n",
       "3250    0\n",
       "3673    0\n",
       "2241    1\n",
       "Name: label, dtype: int64"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train = y_train.apply(lambda x: 1 if x == 'H' else 0)\n",
    "y_test = y_test.apply(lambda x: 1 if x == 'H' else 0)\n",
    "\n",
    "y_test[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training a LogisticRegression using a training set size of 3300\n",
      "Trained model in 0.03265571594238281 seconds\n",
      "Predicted in 0.0017886161804199219 seconds\n",
      "Accyracy on training set: 0.6721212121212121.\n",
      "Predicted in 0.0013942718505859375 seconds\n",
      "Accyracy on test set: 0.6575.\n",
      "\n",
      "Training a SVC using a training set size of 3300\n",
      "Trained model in 0.18880367279052734 seconds\n",
      "Predicted in 0.11607813835144043 seconds\n",
      "Accyracy on training set: 0.6533333333333333.\n",
      "Predicted in 0.015352725982666016 seconds\n",
      "Accyracy on test set: 0.655.\n",
      "\n",
      "Training a XGBClassifier using a training set size of 3300\n",
      "Trained model in 0.14288926124572754 seconds\n",
      "Predicted in 0.00505375862121582 seconds\n",
      "Accyracy on training set: 0.9409090909090909.\n",
      "Predicted in 0.0010502338409423828 seconds\n",
      "Accyracy on test set: 0.6.\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,\n",
       "       0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 1, 0, 1, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,\n",
       "       1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1,\n",
       "       0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,\n",
       "       0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 0,\n",
       "       0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0,\n",
       "       0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1,\n",
       "       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "\n",
    "\n",
    "\n",
    "# This are developer tools, uncomment them and play around if you want to\n",
    "train_predict(clf_A, X_train, y_train, X_test, y_test)\n",
    "print()\n",
    "train_predict(clf_B, X_train, y_train, X_test, y_test)\n",
    "print()\n",
    "train_predict(clf_C, X_train, y_train, X_test, y_test)\n",
    "print()\n",
    "clf_A.fit(X_train, y_train)\n",
    "y_pred = clf_A.predict(X_test)\n",
    "\n",
    "\n",
    "y_pred  \n",
    "#     end = time.time()\n",
    "#     # Print and return results\n",
    "#     print(f\"Predicted in {end - start} seconds\")\n",
    "# #     print(f1_score(target, y_pred, pos_label=1))\n",
    "#     print(y_pred == target)\n",
    "\n",
    "\n",
    "\n",
    "# I will save the model after training:\n",
    "# with open('models/finalized_model.sav', 'w') as f:\n",
    "#     pass\n",
    "\n",
    "# Main training of the model\n",
    "# start = time.time()\n",
    "# print('training home model')\n",
    "# clf_boosted.fit(X_train, y_train)\n",
    "# print(f\"took {time.time() - start}\")\n",
    "# start = time.time()\n",
    "\n",
    "\n",
    "\n",
    "# # save the models to disk\n",
    "# pickle.dump(clf_boosted, open('models/finalized_model.sav', 'wb'))\n",
    "# y_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Методом спроб і помилок я визначив, що під цю проблему найкраще підходить XGBClassifier. Зараз я в лоб переберу вхідні параметри так, щоб максимізувати точність.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Boosted XGBClassifier:\n",
      "Predicted in 0.001989603042602539 seconds\n",
      "Accuracy score for test set: 0.6875757575757576\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "\n",
    "\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import make_scorer\n",
    "import sklearn\n",
    "\n",
    "\n",
    "# Create the parameters list you wish to tune\n",
    "parameters = { 'learning_rate' : [0.1],\n",
    "               'n_estimators' : [40],\n",
    "               'max_depth': [3],\n",
    "               'min_child_weight': [3],\n",
    "               'gamma':[0.4],\n",
    "               'subsample' : [0.8],\n",
    "               'colsample_bytree' : [0.8],\n",
    "               'scale_pos_weight' : [1],\n",
    "               'reg_alpha':[1e-5]\n",
    "             }  \n",
    "\n",
    "best_clf = xgb.XGBClassifier(seed=2)\n",
    "\n",
    "f1_scorer = make_scorer(f1_score,pos_label=1)\n",
    "\n",
    "grid_obj = GridSearchCV(best_clf,\n",
    "                        scoring=f1_scorer,\n",
    "                        param_grid=parameters,\n",
    "                        cv=5)\n",
    "\n",
    "grid_obj = grid_obj.fit(X_train,y_train)\n",
    "\n",
    "# Get the estimator\n",
    "best_clf = grid_obj.best_estimator_\n",
    "\n",
    "# the final score for training and testing after parameter tuning\n",
    "print(f\"Boosted XGBClassifier:\")\n",
    "f1, acc = predict_labels(best_clf, X_train, y_train)\n",
    "print(f\"Accuracy score for test set: {acc}\")\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "source": [
    "### Saving a model we got to disk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# save the models to disk\n",
    "pickle.dump(best_clf, open('models/finalized_model.sav', 'wb'))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Testing output.\n",
    "Splitting again, but this time with 100% of training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[08:29:39] WARNING: /workspace/src/learner.cc:480: \n",
      "Parameters: { scale_pos_weight } might not be used.\n",
      "\n",
      "  This may not be accurate due to some parameters are only used in language bindings but\n",
      "  passed down to XGBoost core.  Or some parameters are not used but slip through this\n",
      "  verification. Please open an issue if you find above cases.\n",
      "\n",
      "\n",
      "      H_goal_diff  A_goal_diff  H_win  A_win  H_win_against  A_win_against\n",
      "3596         38.0          5.0   16.0    7.0            2.0            0.0\n",
      "Away win:  0.09261296\n",
      "Draw:  0.34364447\n",
      "Home win:  0.5637426\n"
     ]
    }
   ],
   "source": [
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(features, labels, \n",
    "                                                    test_size = 10,\n",
    "                                                    random_state = 2,\n",
    "                                                    stratify = labels)\n",
    "\n",
    "\n",
    "# load the model from disk\n",
    "clf_boosted = pickle.load(open('models/finalized_model.sav', 'rb'))\n",
    "clf_boosted.fit(X_train, y_train)\n",
    "# make a prediction\n",
    "\n",
    "to_predict = X_test.iloc[-5:-4]\n",
    "\n",
    "print(to_predict)\n",
    "away_win, draw, home_win = clf_boosted.predict_proba(to_predict)[0]\n",
    "\n",
    "print('Away win: ', away_win)\n",
    "print('Draw: ', draw)\n",
    "print('Home win: ', home_win)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "_change_revision": 2,
  "_is_fork": false,
  "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": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}