notebooks/02-expected-goals-calculate-xg-and-shap.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\ay54372\\Miniconda3\\envs\\expected-goals\\lib\\site-packages\\tqdm\\auto.py:22: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from joblib import load\n",
    "from sklearn.inspection import permutation_importance\n",
    "import shap"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load data and model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "df_lgbm = pd.read_parquet(os.path.join('..', 'data', 'modelling', 'lgbm.parquet'))\n",
    "df_lgbm.sort_index(inplace=True)\n",
    "df_X = df_lgbm.drop(['goal', 'split', 'match_id', 'wyscout_id', 'statsbomb_id'], axis=1).copy()\n",
    "model = load(os.path.join('..', 'models', 'lgbm_model.joblib'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Get feature names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = df_X.columns\n",
    "features = np.array([f.replace('_', ' ') for f in features])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "xg predictions. First have to fit to training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_lgbm['xg'] = model.predict_proba(df_X)[:, 1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Get a datafrane if the uncalibrated/ calibrated probabilities for each of the three models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "models = model.calibrated_classifiers_\n",
    "estimators = [model.base_estimator for model in models]\n",
    "probabilities = []\n",
    "for i in range(3):\n",
    "    probabilities.append(estimators[i].predict_proba(df_X)[:, 1])\n",
    "    probabilities.append(models[i].predict_proba(df_X)[:, 1])\n",
    "df_probabilities = pd.DataFrame(np.vstack(probabilities).T, columns=['uncalibrated0', 'calibrated0',\n",
    "                                                                     'uncalibrated1', 'calibrated1',\n",
    "                                                                     'uncalibrated2', 'calibrated2'])\n",
    "df_probabilities['calibrated'] = model.predict_proba(df_X)[:, 1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Shap values from shap package (estimate as probabilites rather than log odds deviation from bias).\n",
    "Done in loop for the estimators in the calibrated classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|===================| 73842/73958 [03:11<00:00]        "
     ]
    }
   ],
   "source": [
    "# this takes a while, but calculates the contributions as probabilities\n",
    "# https://github.com/slundberg/shap/issues/963\n",
    "sample = data=df_X.sample(500).astype(np.float32)\n",
    "contributions = []\n",
    "for estimator in estimators:\n",
    "    explainer = shap.TreeExplainer(estimator, data=sample, model_output='probability')\n",
    "    shap_values_probability = explainer.shap_values(df_X)\n",
    "    bias = explainer.expected_value\n",
    "    df_contributions_probability = pd.DataFrame(shap_values_probability, columns=features)\n",
    "    df_contributions_probability['bias'] = bias\n",
    "    contributions.append(df_contributions_probability)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Scale the uncalibrated contributions to sum to the calibrated predicted probability"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "scaled_contributions = []\n",
    "for i in range(3):\n",
    "    scaled = (contributions[i]\n",
    "              .divide(df_probabilities[f'uncalibrated{i}'], axis=0)\n",
    "              .multiply(df_probabilities[f'calibrated{i}'], axis=0))\n",
    "    scaled_contributions.append(scaled)   \n",
    "df_scaled_contributions = (scaled_contributions[0] + scaled_contributions[1] + scaled_contributions[2])/3."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Remove a few where the scaled contributions don't match the actual xg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(73958, 26)\n",
      "(73958, 4)\n"
     ]
    }
   ],
   "source": [
    "df_base = df_lgbm[['match_id', 'wyscout_id', 'statsbomb_id', 'xg']]\n",
    "print(df_scaled_contributions.shape)\n",
    "print(df_base.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Where scaled contributions not within 5 decimal places of the Xg value set to missing (some rounding errors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_scaled_contributions.reset_index(drop=True, inplace=True)\n",
    "df_base.reset_index(drop=True, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "mask_contributions_not_equal = ((df_scaled_contributions.sum(axis=1) - df_base.xg).abs().round(5) > 0)\n",
    "df_scaled_contributions[mask_contributions_not_equal] = np.nan"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Return a dataframe with the contirubtions, match and event ids, and the contributions and xg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_base = df_base.merge(df_scaled_contributions,\n",
    "                        left_index=True,\n",
    "                        right_index=True,\n",
    "                        how='left',\n",
    "                        validate='1:1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 73958 entries, 0 to 73957\n",
      "Data columns (total 30 columns):\n",
      " #   Column               Non-Null Count  Dtype  \n",
      "---  ------               --------------  -----  \n",
      " 0   match_id             73958 non-null  int64  \n",
      " 1   wyscout_id           42648 non-null  float64\n",
      " 2   statsbomb_id         31310 non-null  object \n",
      " 3   xg                   73958 non-null  float64\n",
      " 4   shot type name       73958 non-null  float64\n",
      " 5   x                    73958 non-null  float64\n",
      " 6   y                    73958 non-null  float64\n",
      " 7   counter attack       73958 non-null  float64\n",
      " 8   fast break           73958 non-null  float64\n",
      " 9   strong foot          73958 non-null  float64\n",
      " 10  body part name       73958 non-null  float64\n",
      " 11  assist type          73958 non-null  float64\n",
      " 12  pass end y           73958 non-null  float64\n",
      " 13  pass end x           73958 non-null  float64\n",
      " 14  pass switch          73958 non-null  float64\n",
      " 15  pass cross           73958 non-null  float64\n",
      " 16  pass cut back        73958 non-null  float64\n",
      " 17  pass height name     73958 non-null  float64\n",
      " 18  pass technique name  73958 non-null  float64\n",
      " 19  carry length         73958 non-null  float64\n",
      " 20  shot one on one      73958 non-null  float64\n",
      " 21  shot open goal       73958 non-null  float64\n",
      " 22  under pressure       73958 non-null  float64\n",
      " 23  area shot            73958 non-null  float64\n",
      " 24  area goal            73958 non-null  float64\n",
      " 25  n angle              73958 non-null  float64\n",
      " 26  goalkeeper x         73958 non-null  float64\n",
      " 27  goalkeeper y         73958 non-null  float64\n",
      " 28  smart pass           73958 non-null  float64\n",
      " 29  bias                 73958 non-null  float64\n",
      "dtypes: float64(28), int64(1), object(1)\n",
      "memory usage: 16.9+ MB\n"
     ]
    }
   ],
   "source": [
    "df_base.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Save the dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_base.to_parquet(os.path.join('..', 'data', 'modelling', 'xg_shap.parquet'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Use Shap to plot a contribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2000x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "idx=2\n",
    "shap.force_plot(df_scaled_contributions.iloc[idx, -1], \n",
    "                df_scaled_contributions.iloc[idx, :-1].values,\n",
    "                features=df_X.iloc[idx].values,\n",
    "                feature_names=features, matplotlib=True, show=False)\n",
    "fig = plt.gcf()\n",
    "fig.savefig(os.path.join('..', 'figures', '08_shap_example.png'), bbox_inches = 'tight', pad_inches = 0.2)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}