{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# AutoGluon and probability calibration\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 確率補正に必要なsklearn APIの取り付け\n",
    "\n",
    "本連載では分類モデルの予測値を信頼性曲線にプロットしたり、\n",
    "クラス確率に近づける確率補正について取り上げています。\n",
    "前回、sklearn APIが元々備わっていたLightGBMモデルを対象にしましたが、\n",
    "今回は、[AutoGluon Tabular](https://pypi.org/project/autogluon.tabular/) という AutoML を対象に、\n",
    "確率補正に必要なAPIメソッドを取り付ける\n",
    "具体例をご紹介します。\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ライブラリの用意\n",
    "\n",
    "sklearn を使って信頼性曲線を書いたり確率補正します。\n",
    "図形は Matplotlib と Plotly で作ります。\n",
    "補正対象のモデルとしては \n",
    "[AutoGluon Tabular](https://pypi.org/project/autogluon.tabular/)\n",
    "を使います。\n",
    "AutoGluon内部では \n",
    "[lightgbm](https://lightgbm.readthedocs.io/en/latest/),\n",
    "[catboost](https://catboost.ai/),\n",
    "[xgboost](https://xgboost.readthedocs.io/en/latest/)\n",
    "などが使われます。\n",
    "\n",
    "Condaやpipでそれぞれインストールできます。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## サンプルデータ\n",
    "\n",
    "[Adult Census Income (国勢調査の成人収入)](https://archive.ics.uci.edu/ml/datasets/census+income)\n",
    "を使います。5万ドル以上の年収があるかどうかを分類するデータセットです。\n",
    "このデータは元々正例の割合が少なく、偏っていますが、\n",
    "更にノイズを投入しました。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original dataframe shape (32561, 12)\n",
      "Noisy dataframe shape (32561, 1212)\n",
      "Classes [False  True]\n",
      "学習・補正・テスト用データ比率 [0.74997697 0.12284635 0.12717668]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import shap\n",
    "import sklearn\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "## Census income\n",
    "X, y = shap.datasets.adult()\n",
    "X = X.values\n",
    "\n",
    "print(\"Original dataframe shape\", X.shape)\n",
    "n_samples, n_features = X.shape\n",
    "\n",
    "# Add noise\n",
    "random_state = np.random.RandomState(0)\n",
    "X = X + 4 * random_state.randn(n_samples, n_features)\n",
    "X = np.c_[X, random_state.randn(n_samples, 100 * n_features)]\n",
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    X, y, test_size=0.25, random_state=random_state\n",
    ")\n",
    "print(\"Noisy dataframe shape\", X.shape)\n",
    "print(\"Classes\", np.unique(y))\n",
    "\n",
    "X_test, X_calib, y_test, y_calib = train_test_split(\n",
    "    X_test,\n",
    "    y_test,\n",
    "    test_size=4000 / len(X_test),\n",
    "    random_state=random_state,\n",
    ")\n",
    "print(\"学習・補正・テスト用データ比率\", np.array([len(X_train), len(X_calib), len(X_test)]) / len(X))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "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>0</th>\n      <th>1</th>\n      <th>2</th>\n      <th>3</th>\n      <th>4</th>\n      <th>5</th>\n      <th>6</th>\n      <th>7</th>\n      <th>8</th>\n      <th>9</th>\n      <th>...</th>\n      <th>1203</th>\n      <th>1204</th>\n      <th>1205</th>\n      <th>1206</th>\n      <th>1207</th>\n      <th>1208</th>\n      <th>1209</th>\n      <th>1210</th>\n      <th>1211</th>\n      <th>y</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>55.150679</td>\n      <td>0.190716</td>\n      <td>11.639823</td>\n      <td>0.984050</td>\n      <td>4.207856</td>\n      <td>-0.438618</td>\n      <td>4.075127</td>\n      <td>-0.177213</td>\n      <td>4.419674</td>\n      <td>2261.400319</td>\n      <td>...</td>\n      <td>-0.421482</td>\n      <td>-0.492354</td>\n      <td>2.188931</td>\n      <td>1.890532</td>\n      <td>1.197646</td>\n      <td>-0.915642</td>\n      <td>-0.745367</td>\n      <td>1.309365</td>\n      <td>-0.558290</td>\n      <td>False</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>44.324625</td>\n      <td>4.568088</td>\n      <td>14.691406</td>\n      <td>4.085347</td>\n      <td>-0.986166</td>\n      <td>5.562246</td>\n      <td>-0.689750</td>\n      <td>-8.195224</td>\n      <td>1.908657</td>\n      <td>-2.469216</td>\n      <td>...</td>\n      <td>-0.248147</td>\n      <td>0.188788</td>\n      <td>0.759042</td>\n      <td>1.183562</td>\n      <td>0.641865</td>\n      <td>0.095657</td>\n      <td>1.481117</td>\n      <td>-1.294230</td>\n      <td>-1.477307</td>\n      <td>True</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>49.161768</td>\n      <td>3.504227</td>\n      <td>10.344601</td>\n      <td>2.898027</td>\n      <td>18.864769</td>\n      <td>3.734059</td>\n      <td>5.944307</td>\n      <td>0.445238</td>\n      <td>3.814409</td>\n      <td>4.375051</td>\n      <td>...</td>\n      <td>-0.108891</td>\n      <td>-1.597474</td>\n      <td>-0.557484</td>\n      <td>0.274719</td>\n      <td>0.834568</td>\n      <td>-0.155843</td>\n      <td>0.292481</td>\n      <td>0.363087</td>\n      <td>0.687930</td>\n      <td>True</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>38.375623</td>\n      <td>10.634179</td>\n      <td>14.367692</td>\n      <td>11.401001</td>\n      <td>11.490676</td>\n      <td>-1.866726</td>\n      <td>-0.675371</td>\n      <td>8.023515</td>\n      <td>4.178944</td>\n      <td>-7.578039</td>\n      <td>...</td>\n      <td>1.123319</td>\n      <td>-0.363197</td>\n      <td>-0.915543</td>\n      <td>-0.721841</td>\n      <td>1.290744</td>\n      <td>-0.400617</td>\n      <td>-0.254988</td>\n      <td>-1.682269</td>\n      <td>-1.519599</td>\n      <td>False</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>51.941775</td>\n      <td>4.504722</td>\n      <td>8.233571</td>\n      <td>0.416077</td>\n      <td>0.989725</td>\n      <td>-2.964122</td>\n      <td>-0.766150</td>\n      <td>2.836963</td>\n      <td>1.554746</td>\n      <td>-7.509606</td>\n      <td>...</td>\n      <td>0.700531</td>\n      <td>0.465055</td>\n      <td>1.754451</td>\n      <td>-0.510019</td>\n      <td>-0.118499</td>\n      <td>0.867805</td>\n      <td>-0.419116</td>\n      <td>-0.296483</td>\n      <td>0.815914</td>\n      <td>False</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows × 1213 columns</p>\n</div>",
      "text/plain": "           0          1          2          3          4         5         6  \\\n0  55.150679   0.190716  11.639823   0.984050   4.207856 -0.438618  4.075127   \n1  44.324625   4.568088  14.691406   4.085347  -0.986166  5.562246 -0.689750   \n2  49.161768   3.504227  10.344601   2.898027  18.864769  3.734059  5.944307   \n3  38.375623  10.634179  14.367692  11.401001  11.490676 -1.866726 -0.675371   \n4  51.941775   4.504722   8.233571   0.416077   0.989725 -2.964122 -0.766150   \n\n          7         8            9  ...      1203      1204      1205  \\\n0 -0.177213  4.419674  2261.400319  ... -0.421482 -0.492354  2.188931   \n1 -8.195224  1.908657    -2.469216  ... -0.248147  0.188788  0.759042   \n2  0.445238  3.814409     4.375051  ... -0.108891 -1.597474 -0.557484   \n3  8.023515  4.178944    -7.578039  ...  1.123319 -0.363197 -0.915543   \n4  2.836963  1.554746    -7.509606  ...  0.700531  0.465055  1.754451   \n\n       1206      1207      1208      1209      1210      1211      y  \n0  1.890532  1.197646 -0.915642 -0.745367  1.309365 -0.558290  False  \n1  1.183562  0.641865  0.095657  1.481117 -1.294230 -1.477307   True  \n2  0.274719  0.834568 -0.155843  0.292481  0.363087  0.687930   True  \n3 -0.721841  1.290744 -0.400617 -0.254988 -1.682269 -1.519599  False  \n4 -0.510019 -0.118499  0.867805 -0.419116 -0.296483  0.815914  False  \n\n[5 rows x 1213 columns]"
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from autogluon.tabular import TabularDataset\n",
    "\n",
    "tr_data = TabularDataset(X_train)\n",
    "tr_data[\"y\"] = y_train\n",
    "tr_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "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>0</th>\n      <th>1</th>\n      <th>2</th>\n      <th>3</th>\n      <th>4</th>\n      <th>5</th>\n      <th>6</th>\n      <th>7</th>\n      <th>8</th>\n      <th>9</th>\n      <th>...</th>\n      <th>1203</th>\n      <th>1204</th>\n      <th>1205</th>\n      <th>1206</th>\n      <th>1207</th>\n      <th>1208</th>\n      <th>1209</th>\n      <th>1210</th>\n      <th>1211</th>\n      <th>y</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>62.255383</td>\n      <td>5.831143</td>\n      <td>12.638373</td>\n      <td>-4.375959</td>\n      <td>11.950415</td>\n      <td>2.638438</td>\n      <td>-2.181352</td>\n      <td>-1.075052</td>\n      <td>-1.040339</td>\n      <td>-4.046030</td>\n      <td>...</td>\n      <td>0.627686</td>\n      <td>-1.381202</td>\n      <td>0.235047</td>\n      <td>0.781438</td>\n      <td>0.352366</td>\n      <td>0.583754</td>\n      <td>1.224902</td>\n      <td>0.350085</td>\n      <td>0.407340</td>\n      <td>False</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>42.304418</td>\n      <td>1.507765</td>\n      <td>9.225236</td>\n      <td>7.568787</td>\n      <td>15.296958</td>\n      <td>4.998232</td>\n      <td>3.006109</td>\n      <td>0.383081</td>\n      <td>5.057561</td>\n      <td>3.583399</td>\n      <td>...</td>\n      <td>-1.687568</td>\n      <td>0.528663</td>\n      <td>-1.178646</td>\n      <td>1.333547</td>\n      <td>-0.557322</td>\n      <td>0.489001</td>\n      <td>0.372340</td>\n      <td>0.810345</td>\n      <td>-0.431861</td>\n      <td>False</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>25.362705</td>\n      <td>0.099063</td>\n      <td>4.370707</td>\n      <td>17.376790</td>\n      <td>9.849307</td>\n      <td>0.636375</td>\n      <td>7.183748</td>\n      <td>-0.959921</td>\n      <td>0.136696</td>\n      <td>4.054935</td>\n      <td>...</td>\n      <td>-0.892634</td>\n      <td>0.120163</td>\n      <td>0.872914</td>\n      <td>0.651086</td>\n      <td>0.625578</td>\n      <td>-1.243911</td>\n      <td>0.635322</td>\n      <td>1.444986</td>\n      <td>0.077606</td>\n      <td>False</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>10.007130</td>\n      <td>12.341750</td>\n      <td>4.399064</td>\n      <td>-1.052642</td>\n      <td>10.981154</td>\n      <td>-1.703968</td>\n      <td>3.493526</td>\n      <td>-0.566090</td>\n      <td>-1.110927</td>\n      <td>7.241747</td>\n      <td>...</td>\n      <td>-1.264475</td>\n      <td>-2.235201</td>\n      <td>0.915248</td>\n      <td>1.041231</td>\n      <td>0.544913</td>\n      <td>1.468259</td>\n      <td>-1.761762</td>\n      <td>3.205364</td>\n      <td>0.742915</td>\n      <td>False</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>25.622787</td>\n      <td>4.829471</td>\n      <td>7.509705</td>\n      <td>3.995892</td>\n      <td>5.798364</td>\n      <td>2.085136</td>\n      <td>0.069773</td>\n      <td>-0.438765</td>\n      <td>-4.660486</td>\n      <td>1.304545</td>\n      <td>...</td>\n      <td>0.425200</td>\n      <td>1.008140</td>\n      <td>0.618921</td>\n      <td>1.234493</td>\n      <td>-0.278199</td>\n      <td>0.851198</td>\n      <td>0.133332</td>\n      <td>-0.467102</td>\n      <td>1.612342</td>\n      <td>False</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows × 1213 columns</p>\n</div>",
      "text/plain": "           0          1          2          3          4         5         6  \\\n0  62.255383   5.831143  12.638373  -4.375959  11.950415  2.638438 -2.181352   \n1  42.304418   1.507765   9.225236   7.568787  15.296958  4.998232  3.006109   \n2  25.362705   0.099063   4.370707  17.376790   9.849307  0.636375  7.183748   \n3  10.007130  12.341750   4.399064  -1.052642  10.981154 -1.703968  3.493526   \n4  25.622787   4.829471   7.509705   3.995892   5.798364  2.085136  0.069773   \n\n          7         8         9  ...      1203      1204      1205      1206  \\\n0 -1.075052 -1.040339 -4.046030  ...  0.627686 -1.381202  0.235047  0.781438   \n1  0.383081  5.057561  3.583399  ... -1.687568  0.528663 -1.178646  1.333547   \n2 -0.959921  0.136696  4.054935  ... -0.892634  0.120163  0.872914  0.651086   \n3 -0.566090 -1.110927  7.241747  ... -1.264475 -2.235201  0.915248  1.041231   \n4 -0.438765 -4.660486  1.304545  ...  0.425200  1.008140  0.618921  1.234493   \n\n       1207      1208      1209      1210      1211      y  \n0  0.352366  0.583754  1.224902  0.350085  0.407340  False  \n1 -0.557322  0.489001  0.372340  0.810345 -0.431861  False  \n2  0.625578 -1.243911  0.635322  1.444986  0.077606  False  \n3  0.544913  1.468259 -1.761762  3.205364  0.742915  False  \n4 -0.278199  0.851198  0.133332 -0.467102  1.612342  False  \n\n[5 rows x 1213 columns]"
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "te_data = TabularDataset(X_test)\n",
    "te_data[\"y\"] = y_test\n",
    "te_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": "count     24420\nunique        2\ntop       False\nfreq      18577\nName: y, dtype: object"
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tr_data.y.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": "count      4141\nunique        2\ntop       False\nfreq       3098\nName: y, dtype: object"
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "te_data.y.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## モデル学習"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "AutoGluonモデルを学習データにフィットさせます。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Warning: path already exists! This predictor may overwrite an existing predictor! path=\"trained-model\"\n",
      "Beginning AutoGluon training ... Time limit = 30s\n",
      "AutoGluon will save models to \"trained-model/\"\n",
      "AutoGluon Version:  0.1.0\n",
      "Train Data Rows:    24420\n",
      "Train Data Columns: 1212\n",
      "Preprocessing data ...\n",
      "AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).\n",
      "\t2 unique label values:  [False, True]\n",
      "\tIf 'binary' is not the correct problem_type, please manually specify the problem_type argument in fit() (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])\n",
      "Selected class <--> label mapping:  class 1 = True, class 0 = False\n",
      "Using Feature Generators to preprocess the data ...\n",
      "Fitting AutoMLPipelineFeatureGenerator...\n",
      "\tAvailable Memory:                    6459.76 MB\n",
      "\tTrain Data (Original)  Memory Usage: 236.78 MB (3.7% of available memory)\n",
      "\tInferring data type of each feature based on column values. Set feature_metadata_in to manually specify special dtypes of the features.\n",
      "\tStage 1 Generators:\n",
      "\t\tFitting AsTypeFeatureGenerator...\n",
      "\tStage 2 Generators:\n",
      "\t\tFitting FillNaFeatureGenerator...\n",
      "\tStage 3 Generators:\n",
      "\t\tFitting IdentityFeatureGenerator...\n",
      "\tStage 4 Generators:\n",
      "\t\tFitting DropUniqueFeatureGenerator...\n",
      "\tTypes of features in original data (raw dtype, special dtypes):\n",
      "\t\t('float', []) : 1212 | ['0', '1', '2', '3', '4', ...]\n",
      "\tTypes of features in processed data (raw dtype, special dtypes):\n",
      "\t\t('float', []) : 1212 | ['0', '1', '2', '3', '4', ...]\n",
      "\t4.5s = Fit runtime\n",
      "\t1212 features in original data used to generate 1212 features in processed data.\n",
      "\tTrain Data (Processed) Memory Usage: 236.78 MB (3.4% of available memory)\n",
      "Data preprocessing and feature engineering runtime = 5.39s ...\n",
      "AutoGluon will gauge predictive performance using evaluation metric: 'accuracy'\n",
      "\tTo change this, specify the eval_metric argument of fit()\n",
      "Automatically generating train/validation split with holdout_frac=0.1, Train Rows: 21978, Val Rows: 2442\n",
      "Fitting model: LightGBM ... Training model for up to 24.61s of the 24.6s of remaining time.\n",
      "\t0.8084\t = Validation accuracy score\n",
      "\t3.28s\t = Training runtime\n",
      "\t0.05s\t = Validation runtime\n",
      "Fitting model: CatBoost ... Training model for up to 21.28s of the 21.27s of remaining time.\n",
      "\t0.8157\t = Validation accuracy score\n",
      "\t1.74s\t = Training runtime\n",
      "\t0.02s\t = Validation runtime\n",
      "Fitting model: XGBoost ... Training model for up to 19.51s of the 19.5s of remaining time.\n",
      "ntree_limit is deprecated, use `iteration_range` or model slicing instead.\n",
      "\t0.8215\t = Validation accuracy score\n",
      "\t16.77s\t = Training runtime\n",
      "\t0.24s\t = Validation runtime\n",
      "Fitting model: NeuralNetMXNet ... Training model for up to 2.47s of the 2.46s of remaining time.\n",
      "\tTime limit exceeded... Skipping NeuralNetMXNet.\n",
      "ntree_limit is deprecated, use `iteration_range` or model slicing instead.\n",
      "Fitting model: WeightedEnsemble_L2 ... Training model for up to 24.61s of the -5.42s of remaining time.\n",
      "\t0.8219\t = Validation accuracy score\n",
      "\t0.35s\t = Training runtime\n",
      "\t0.0s\t = Validation runtime\n",
      "AutoGluon training complete, total runtime = 36.23s ...\n",
      "TabularPredictor saved. To load, use: predictor = TabularPredictor.load(\"trained-model/\")\n"
     ]
    }
   ],
   "source": [
    "from autogluon.tabular import TabularPredictor\n",
    "\n",
    "save_path = \"trained-model\"\n",
    "predictor = TabularPredictor(label=\"y\", path=save_path).fit(\n",
    "    tr_data, hyperparameters=\"toy\", time_limit=30\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Sklearnラッパー\n",
    "\n",
    "sklearnを使って確率補正を行うために、必要なインタフェースを用意する。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "試験データでメトリック AUC, Brier を図り、信頼性曲線を書いてみます。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastcore.basics import store_attr\n",
    "\n",
    "\n",
    "class AutoGluonWrapper:\n",
    "    \"\"\"\n",
    "    sklearnを使って信頼性曲線を描いたり、確率補正を行うために、\n",
    "    必要なインタフェースを用意する。\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(\n",
    "        self,\n",
    "        trained_model_path,  # AutoGluon学習済みモデルの保存パス\n",
    "        classes_,  # sklearn APIに求められる属性\n",
    "    ):\n",
    "        store_attr()\n",
    "\n",
    "    def load_model(self):\n",
    "        \"\"\" AutoGluon学習済みモデルをロード \"\"\"\n",
    "        self.ag_model = TabularPredictor.load(self.trained_model_path)\n",
    "\n",
    "    def fit(self):\n",
    "        \"\"\" sklearn API に求められるメソッド \"\"\"\n",
    "        return True\n",
    "\n",
    "    def predict_proba(self, X):\n",
    "        \"\"\" sklearn API に求められるメソッド \"\"\"\n",
    "        X = TabularDataset(X)\n",
    "        proba = self.ag_model.predict_proba(X)\n",
    "        return proba.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ag_ = AutoGluonWrapper(save_path, classes_=np.unique(y))\n",
    "ag_.load_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 信頼性曲線"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "試験データに対してメトリック AUC, Brier を図り、信頼性曲線を書いてみます。\n",
    "前回定義したメソッド `plot_calibration_curve` を使います。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from kowaza.proba_calib import plot_calibration_curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ntree_limit is deprecated, use `iteration_range` or model slicing instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AutoGluon:\n",
      "\tAUC  : 0.824\n",
      "\tBrier: 0.153\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": "<Figure size 720x720 with 2 Axes>"
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_calibration_curve(dict(AutoGluon=ag_), X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 確率補正の実施\n",
    "\n",
    "補正するために[sklearn.calibration. CalibratedClassifierCV](https://scikit-learn.org/stable/modules/generated/sklearn.calibration.CalibratedClassifierCV.html#sklearn.calibration.CalibratedClassifierCV)を使います。 `cv = \"prefit\"` と指定することによって、ベースモデルが学習済みであり、補正モデルに渡すデータは全量補正用であることを伝えます。\n",
    "\n",
    "今回は確率補正方法として `sigmoid` と `isotonic` をそれぞれ使ってみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ntree_limit is deprecated, use `iteration_range` or model slicing instead.\n"
     ]
    },
    {
     "data": {
      "text/plain": "CalibratedClassifierCV(base_estimator=<__main__.AutoGluonWrapper object at 0x7fbb9497d7c0>,\n                       cv='prefit')"
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.calibration import CalibratedClassifierCV, calibration_curve\n",
    "\n",
    "sigmoid = CalibratedClassifierCV(ag_, cv=\"prefit\", method=\"sigmoid\")\n",
    "sigmoid.fit(X_calib, y_calib)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ntree_limit is deprecated, use `iteration_range` or model slicing instead.\n"
     ]
    },
    {
     "data": {
      "text/plain": "CalibratedClassifierCV(base_estimator=<__main__.AutoGluonWrapper object at 0x7fbb9497d7c0>,\n                       cv='prefit', method='isotonic')"
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "isotonic = CalibratedClassifierCV(ag_, cv=\"prefit\", method=\"isotonic\")\n",
    "isotonic.fit(X_calib, y_calib)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "補正前と補正後のモデルの信頼性曲線を同じ図に書いて比較してみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ntree_limit is deprecated, use `iteration_range` or model slicing instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AutoGluon:\n",
      "\tAUC  : 0.824\n",
      "\tBrier: 0.153\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ntree_limit is deprecated, use `iteration_range` or model slicing instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sigmoid:\n",
      "\tAUC  : 0.824\n",
      "\tBrier: 0.131\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ntree_limit is deprecated, use `iteration_range` or model slicing instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Isotonic:\n",
      "\tAUC  : 0.824\n",
      "\tBrier: 0.131\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": "<Figure size 720x720 with 2 Axes>"
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_calibration_curve(\n",
    "    dict(\n",
    "        AutoGluon=ag_,\n",
    "        Sigmoid=sigmoid,\n",
    "        Isotonic=isotonic,\n",
    "    ),\n",
    "    X_test,\n",
    "    y_test,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "補正の結果、信頼性曲線が改善され、Brierスコアも改善されました。\n",
    "最も、対象データセットにノイズを投入せずにAutoGluonに学習させると、確率補正を行わなくてもきれいな信頼性曲線が得られます。\n",
    "しかし、実運用では、データに必ずノイズが含まれるし、データの分布も時間とともに少しずつ変化していくものなので、信頼性曲線をプロットして必要に応じて確率補正を行う必要があります。\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## まとめ\n",
    "\n",
    "第１回目に、分類モデルの出力値が必ずしもクラス確率とは限らないので、信頼性曲線を確認したり、確率補正を行ってみました。\n",
    "今回は、sklearnが規定しているインタフェースを持たない学習済みモデルの補正実験を行いました。\n",
    "次回は、確率補正に関する背景的な理論について書ければと思います。\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "kowaza",
   "language": "python",
   "name": "kowaza"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}