2_nba_players_modelisation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 2,
  "metadata": {
    "colab": {
      "name": "nba_players_modelisation.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "# Modélisation\r\n",
        "\r\n",
        "Ce NoteBook propose une modélisation pour les données analysées dans le NoteBook précédent (nba_players_analyse.ipynb).\r\n",
        "\r\n",
        "Ce NoteBook est composé des parties suivantes :\r\n",
        "- Accès aux données stockées sur Google Drive,\r\n",
        "- Import des librairies,\r\n",
        "- Configuration du NoteBook,\r\n",
        "- Chargement des données,\r\n",
        "- Séparation du jeu de données,\r\n",
        "- Transformation des variables,\r\n",
        "- Modèle baseline,\r\n",
        "- Comparaison de modèles,\r\n",
        "- Meilleur modèle,\r\n",
        "- Problème avec API Rest."
      ],
      "metadata": {
        "id": "20KiGQx8ip3S"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Accès aux données stockées sur Google Drive"
      ],
      "metadata": {
        "id": "Pw7Ac2Ygjd6L"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "source": [
        "from google.colab import drive\r\n",
        "drive.mount(\"/content/drive\")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3G4Av4AOiCUw",
        "outputId": "8a9c5945-d5dc-4f70-a305-52eef3fd4187"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Import des librairies"
      ],
      "metadata": {
        "id": "HnutEXffmcwQ"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "source": [
        "# skopt: model optimization lib\r\n",
        "!pip install -q scikit-optimize"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[?25l\r\u001b[K     |███▎                            | 10 kB 20.4 MB/s eta 0:00:01\r\u001b[K     |██████▌                         | 20 kB 26.8 MB/s eta 0:00:01\r\u001b[K     |█████████▊                      | 30 kB 31.9 MB/s eta 0:00:01\r\u001b[K     |█████████████                   | 40 kB 33.4 MB/s eta 0:00:01\r\u001b[K     |████████████████▏               | 51 kB 35.2 MB/s eta 0:00:01\r\u001b[K     |███████████████████▍            | 61 kB 37.8 MB/s eta 0:00:01\r\u001b[K     |██████████████████████▊         | 71 kB 26.5 MB/s eta 0:00:01\r\u001b[K     |██████████████████████████      | 81 kB 27.8 MB/s eta 0:00:01\r\u001b[K     |█████████████████████████████▏  | 92 kB 27.1 MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 101 kB 8.2 MB/s \n",
            "\u001b[?25h"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "G5ABbKUz5KfI",
        "outputId": "dd067e9e-a228-4052-d1a3-97be283fb5e2"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "source": [
        "import time\r\n",
        "\r\n",
        "import pandas as pd\r\n",
        "import numpy as np\r\n",
        "import matplotlib.pyplot as plt\r\n",
        "import seaborn as sns\r\n",
        "\r\n",
        "from sklearn.model_selection import train_test_split, cross_validate\r\n",
        "from sklearn.preprocessing import StandardScaler\r\n",
        "from sklearn.pipeline import Pipeline\r\n",
        "from sklearn.metrics import (\r\n",
        "    f1_score,\r\n",
        "    precision_score,\r\n",
        "    recall_score,\r\n",
        "    confusion_matrix\r\n",
        ")\r\n",
        "from sklearn.dummy import DummyClassifier\r\n",
        "from sklearn.linear_model import LogisticRegression\r\n",
        "from sklearn.svm import SVC\r\n",
        "\r\n",
        "from skopt import BayesSearchCV\r\n",
        "from skopt.space import Real, Integer\r\n",
        "\r\n",
        "import dill as pickle"
      ],
      "outputs": [],
      "metadata": {
        "id": "weITbeTkmfAw"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Configuration du NoteBook"
      ],
      "metadata": {
        "id": "4G-qYvdIreym"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "source": [
        "CONFIG = dict (\r\n",
        "    seed = 0,\r\n",
        "    nbr_folds = 5,\r\n",
        "    n_iter = 66\r\n",
        ")"
      ],
      "outputs": [],
      "metadata": {
        "id": "YTXpCVvTrhsz"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Chargement des données"
      ],
      "metadata": {
        "id": "B8orlP1WmfKm"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "source": [
        "# Chargement des données\n",
        "df = pd.read_csv(\"/content/drive/MyDrive/Gaetan_Travail/ML/df_a_utiliser.csv\")\n",
        "\n",
        "# Vérifions que les données se soient bien chargées\n",
        "print(f\"df shape: {df.shape}\\n\")\n",
        "df.head()"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "df shape: (1273, 12)\n",
            "\n"
          ]
        },
        {
          "output_type": "execute_result",
          "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>GP</th>\n",
              "      <th>PTS</th>\n",
              "      <th>FG%</th>\n",
              "      <th>3P%</th>\n",
              "      <th>FT%</th>\n",
              "      <th>OREB</th>\n",
              "      <th>DREB</th>\n",
              "      <th>AST</th>\n",
              "      <th>STL</th>\n",
              "      <th>BLK</th>\n",
              "      <th>TOV</th>\n",
              "      <th>TARGET_5Yrs</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>36</td>\n",
              "      <td>2.128232</td>\n",
              "      <td>34.7</td>\n",
              "      <td>25.0</td>\n",
              "      <td>69.9</td>\n",
              "      <td>0.530628</td>\n",
              "      <td>1.481605</td>\n",
              "      <td>1.064711</td>\n",
              "      <td>0.336472</td>\n",
              "      <td>0.336472</td>\n",
              "      <td>0.832909</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>35</td>\n",
              "      <td>2.104134</td>\n",
              "      <td>29.6</td>\n",
              "      <td>23.5</td>\n",
              "      <td>76.5</td>\n",
              "      <td>0.405465</td>\n",
              "      <td>1.098612</td>\n",
              "      <td>1.547563</td>\n",
              "      <td>0.741937</td>\n",
              "      <td>0.405465</td>\n",
              "      <td>0.955511</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>74</td>\n",
              "      <td>1.824549</td>\n",
              "      <td>42.2</td>\n",
              "      <td>24.4</td>\n",
              "      <td>67.0</td>\n",
              "      <td>0.405465</td>\n",
              "      <td>0.993252</td>\n",
              "      <td>0.693147</td>\n",
              "      <td>0.405465</td>\n",
              "      <td>0.262364</td>\n",
              "      <td>0.693147</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>58</td>\n",
              "      <td>1.902108</td>\n",
              "      <td>42.6</td>\n",
              "      <td>22.6</td>\n",
              "      <td>68.9</td>\n",
              "      <td>0.693147</td>\n",
              "      <td>0.641854</td>\n",
              "      <td>0.587787</td>\n",
              "      <td>0.470004</td>\n",
              "      <td>0.095310</td>\n",
              "      <td>0.693147</td>\n",
              "      <td>1.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>48</td>\n",
              "      <td>1.704748</td>\n",
              "      <td>52.4</td>\n",
              "      <td>0.0</td>\n",
              "      <td>67.4</td>\n",
              "      <td>0.693147</td>\n",
              "      <td>0.916291</td>\n",
              "      <td>0.262364</td>\n",
              "      <td>0.262364</td>\n",
              "      <td>0.336472</td>\n",
              "      <td>0.587787</td>\n",
              "      <td>1.0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   GP       PTS   FG%   3P%  ...       STL       BLK       TOV  TARGET_5Yrs\n",
              "0  36  2.128232  34.7  25.0  ...  0.336472  0.336472  0.832909          0.0\n",
              "1  35  2.104134  29.6  23.5  ...  0.741937  0.405465  0.955511          0.0\n",
              "2  74  1.824549  42.2  24.4  ...  0.405465  0.262364  0.693147          0.0\n",
              "3  58  1.902108  42.6  22.6  ...  0.470004  0.095310  0.693147          1.0\n",
              "4  48  1.704748  52.4   0.0  ...  0.262364  0.336472  0.587787          1.0\n",
              "\n",
              "[5 rows x 12 columns]"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        },
        "id": "3dku8Ocajj3V",
        "outputId": "6b27467b-2617-41ac-f570-516470d05036"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Séparation du jeu de données\n",
        "\n",
        "Puisqu'il y a un déséquilibre des classes, nous allons utilier la fonction \"train_test_split\" en indiquant une stratification, afin de créer un jeu d'entraînement et un jeu de test.\n",
        "\n",
        "Puis, avec le jeu d'entraînement, nous allons utiliser la fonction StratifiedKFold afin de créer 5 plis (choix arbitraire) qui seront utilisés afin de mettre en place une validation croisée (amélioration de la généralisation d'un modèle)."
      ],
      "metadata": {
        "id": "s2LPI-U-neHn"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "source": [
        "# Creation des jeux d'entraînement et de test\n",
        "\n",
        "# conversion de la cible en int\n",
        "df.TARGET_5Yrs = [int(x) for x in df.TARGET_5Yrs]\n",
        "\n",
        "# séparation du jeu de données\n",
        "train_df, test_df = train_test_split(\n",
        "    df,\n",
        "    test_size=0.2,\n",
        "    random_state=CONFIG[\"seed\"],\n",
        "    stratify=df.TARGET_5Yrs\n",
        ")\n",
        "\n",
        "train_df.reset_index(drop=True, inplace=True)\n",
        "test_df.reset_index(drop=True, inplace=True)\n",
        "\n",
        "print(f\"train set shape: {train_df.shape}\")\n",
        "print(f\"test set shape:  {test_df.shape}\")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "train set shape: (1018, 12)\n",
            "test set shape:  (255, 12)\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CEoAvpivjj5j",
        "outputId": "371116f9-0303-47e8-e8bd-afaad80c00d4"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Transformation des variables\n",
        "\n",
        "Nous allons mettre en place un pipeline afin d'opérer une transformation sur les données n'appartenant qu'au jeu d'entraînement (et ainsi éviter d'utiliser le jeu de validation qui correspondra au pli non utilisé dans le jeu d'entraînement).\n",
        "\n",
        "La transformation utilisée sera \"StandardScaler\"."
      ],
      "metadata": {
        "id": "nWoetmyZsW6_"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Modèle baseline\n",
        "\n",
        "Notre baseline sera un modèle de type \"Dummy Classifier\"."
      ],
      "metadata": {
        "id": "SgwTaao0wXV9"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "#On cree un pipeline de traitement integrant la preparation et le modele Lasso:\n",
        "baseline = Pipeline([\n",
        "  (\"stdScaler\", StandardScaler()),\n",
        "  (\"dummyClf\", DummyClassifier(strategy=\"stratified\"))\n",
        "])\n",
        "\n",
        "#----------------------#\n",
        "\n",
        "# Entraînement du modèle\n",
        "cv_results = cross_validate(\n",
        "    baseline,\n",
        "    train_df.drop(columns=[\"TARGET_5Yrs\"]),\n",
        "    train_df.TARGET_5Yrs,\n",
        "    scoring=\"recall\",\n",
        "    cv=CONFIG[\"nbr_folds\"],\n",
        "    return_estimator=True\n",
        ")\n",
        "\n",
        "#----------------------#\n",
        "\n",
        "# prédictions\n",
        "y_true = test_df.TARGET_5Yrs\n",
        "start = time.time()\n",
        "y_pred = cv_results[\"estimator\"][np.argmax(cv_results[\"test_score\"])].predict(\n",
        "    test_df.drop(columns=[\"TARGET_5Yrs\"])\n",
        ")\n",
        "stop = time.time()\n",
        "\n",
        "print(f\"Recall    = {recall_score(y_true, y_pred):.4f}\")\n",
        "print(f\"F1 score  = {f1_score(y_true, y_pred):.4f}\")\n",
        "print(f\"Precision = {precision_score(y_true, y_pred):.4f}\")\n",
        "print(f\"Temps de prédictions = {stop - start}\")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Recall    = 0.5597\n",
            "F1 score  = 0.5761\n",
            "Precision = 0.5933\n",
            "Temps de prédictions = 0.0021622180938720703\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PVKoIUmeibWg",
        "outputId": "4ba6aa5d-0d90-47d1-8b0e-449732942bc3"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Nous utiliserons comme score baseline :\n",
        "\n",
        "Recall = 0.5597"
      ],
      "metadata": {
        "id": "tim6bUMa3Stp"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Comparaison de modèles\n",
        "\n",
        "Pour optimiser les modèles, je vais utiliser la librairie scikit-optimize. Cette dernière permet d'effectuer des recherches bayésiennes.\n",
        "\n",
        "Pour optimiser le score, j'utilise la métrique \"F1-score\".\n",
        "Celle-ci étant une moyenne harmonique de la précision et du rappel, elle permet d'éviter que le modèle ne fasse des \"erreurs volontaires\" (classer tous les individus en 1 ou 0) pour maximiser la précision ou le rappel.\n",
        "\n",
        "Une fois, l'entraînement fait en tentant d'optimiser le F1-score, nous allons comparer les scores des différents modèles (F1-score et Recall), ainsi que le temps mis pour prédire le jeu de test."
      ],
      "metadata": {
        "id": "QM3mo75n3iyQ"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "source": [
        "# Imbalance du jeu de données\n",
        "\n",
        "def class_weights_function(df_target):\n",
        "  # Estimate class weights based on the imbalance of the data set\n",
        "\n",
        "  neg = df_target.value_counts()[0]\n",
        "  pos = df_target.value_counts()[1]\n",
        "\n",
        "  # weigts to correct imbalance:\n",
        "  total = neg + pos\n",
        "\n",
        "  # Scaling by total/2 helps keep the loss to a similar magnitude.\n",
        "  # The sum of the weights of all examples stays the same.\n",
        "  weight_for_0 = (1 / neg) * (total / 2.0)\n",
        "  weight_for_1 = (1 / pos) * (total / 2.0)\n",
        "\n",
        "  class_weights = {0: weight_for_0, 1: weight_for_1}\n",
        "\n",
        "  return class_weights"
      ],
      "outputs": [],
      "metadata": {
        "id": "G5UVksiTdlOA"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "source": [
        "# Maximisation du F1-score\n",
        "def max_scores(target, y_pred, min_range=30, max_range=70, average_=\"binary\"):\n",
        "\n",
        "    result_threshold = []\n",
        "    result_f1 = []\n",
        "    result_jaccard = []\n",
        "    result_prec = []\n",
        "    result_recall = []\n",
        "\n",
        "    for i in range(min_range, max_range, 1):\n",
        "      threshold = i/100\n",
        "\n",
        "      y_pred_threshold = [1 if x >= threshold else 0 for x in y_pred]\n",
        "                                                           \n",
        "      result_threshold.append(threshold)\n",
        "      result_f1.append(f1_score(target, y_pred_threshold))        \n",
        "      result_prec.append(precision_score(\n",
        "          target,\n",
        "          y_pred_threshold,\n",
        "          average=average_)\n",
        "      )\n",
        "      result_recall.append(recall_score(\n",
        "          target,\n",
        "          y_pred_threshold,\n",
        "          average=average_)\n",
        "      )\n",
        "\n",
        "    # we want to maximize the F1 score:\n",
        "    index = result_f1.index(max(result_f1))\n",
        "    \n",
        "    print(f\"Threshold = {result_threshold[index]}\")\n",
        "    print(\"\\nScores:\\n\")\n",
        "    print(f\"F1 score         = {result_f1[index]:.4f}\")\n",
        "    print(f\"Precision score  = {result_prec[index]:.4f}\")\n",
        "    print(f\"Recall score     = {result_recall[index]:.4f}\")\n",
        "  \n",
        "    return result_threshold[index]"
      ],
      "outputs": [],
      "metadata": {
        "id": "0i2DSjICnw5z"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "source": [
        "from sklearn.metrics import fbeta_score\n",
        "\n",
        "# Maximisation du Fbeta-score\n",
        "def max_scores_fb(target, y_pred, min_range=30, max_range=70, beta_=0.5):\n",
        "\n",
        "    result_threshold = []\n",
        "    result_fb = []\n",
        "\n",
        "    for i in range(min_range, max_range, 1):\n",
        "      threshold = i/100\n",
        "\n",
        "      y_pred_threshold = [1 if x >= threshold else 0 for x in y_pred]\n",
        "                                                           \n",
        "      result_threshold.append(threshold)\n",
        "      result_fb.append(fbeta_score(target, y_pred_threshold, beta=beta_))\n",
        "\n",
        "    # we want to maximize the F1 score:\n",
        "    index = result_fb.index(max(result_fb))\n",
        "    \n",
        "    print(f\"Threshold = {result_threshold[index]}\")\n",
        "    print(\"\\nScores:\\n\")\n",
        "    print(f\"fbeta_score = {result_fb[index]:.4f}\")\n",
        "  \n",
        "    return result_threshold[index]"
      ],
      "outputs": [],
      "metadata": {
        "id": "Ux5r-Q_l9n9Y"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "**Régression logistique**"
      ],
      "metadata": {
        "id": "mU-sv96lTJAK"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "source": [
        "# On cree un pipeline de traitement integrant la preparation et le modele logReg:\n",
        "modele_log_reg = Pipeline([\n",
        "  (\"stdScaler\", StandardScaler()),\n",
        "  (\"logReg\", LogisticRegression(\n",
        "      penalty = \"elasticnet\",\n",
        "      solver = \"saga\",\n",
        "      class_weight = class_weights_function(train_df.TARGET_5Yrs),\n",
        "      random_state = CONFIG[\"seed\"]\n",
        "  ))\n",
        "])\n",
        "\n",
        "# On definit les hyperparametres a faire varier:\n",
        "param_grid = {\n",
        "    \"logReg__C\": Real(1e-5, 1e+5, prior=\"log-uniform\"),\n",
        "    \"logReg__l1_ratio\": Real(0, 1)\n",
        "}\n",
        "\n",
        "# On cree le modele utilisant une recherche bayesienne\n",
        "opt_logReg = BayesSearchCV(\n",
        "    modele_log_reg,\n",
        "    param_grid,\n",
        "    scoring=\"f1\",\n",
        "    cv=CONFIG[\"nbr_folds\"],\n",
        "    n_jobs=-1,\n",
        "    n_iter=CONFIG[\"n_iter\"]\n",
        ")\n",
        "\n",
        "# On entraine le modele\n",
        "opt_logReg.fit(\n",
        "  train_df.drop(columns=[\"TARGET_5Yrs\"]),\n",
        "  train_df.TARGET_5Yrs\n",
        ")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/skopt/optimizer/optimizer.py:449: UserWarning: The objective has been evaluated at this point before.\n",
            "  warnings.warn(\"The objective has been evaluated \"\n",
            "/usr/local/lib/python3.7/dist-packages/skopt/optimizer/optimizer.py:449: UserWarning: The objective has been evaluated at this point before.\n",
            "  warnings.warn(\"The objective has been evaluated \"\n",
            "/usr/local/lib/python3.7/dist-packages/skopt/optimizer/optimizer.py:449: UserWarning: The objective has been evaluated at this point before.\n",
            "  warnings.warn(\"The objective has been evaluated \"\n",
            "/usr/local/lib/python3.7/dist-packages/skopt/optimizer/optimizer.py:449: UserWarning: The objective has been evaluated at this point before.\n",
            "  warnings.warn(\"The objective has been evaluated \"\n",
            "/usr/local/lib/python3.7/dist-packages/skopt/optimizer/optimizer.py:449: UserWarning: The objective has been evaluated at this point before.\n",
            "  warnings.warn(\"The objective has been evaluated \"\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "BayesSearchCV(cv=5, error_score='raise',\n",
              "              estimator=Pipeline(memory=None,\n",
              "                                 steps=[('stdScaler',\n",
              "                                         StandardScaler(copy=True,\n",
              "                                                        with_mean=True,\n",
              "                                                        with_std=True)),\n",
              "                                        ('logReg',\n",
              "                                         LogisticRegression(C=1.0,\n",
              "                                                            class_weight={0: 1.3220779220779222,\n",
              "                                                                          1: 0.8041074249605055},\n",
              "                                                            dual=False,\n",
              "                                                            fit_intercept=True,\n",
              "                                                            intercept_scaling=1,\n",
              "                                                            l1_ratio=None,\n",
              "                                                            max_iter=100,\n",
              "                                                            multi_class='auto',\n",
              "                                                            n_jobs=None,\n",
              "                                                            pena...\n",
              "              fit_params=None, iid=True, n_iter=66, n_jobs=-1, n_points=1,\n",
              "              optimizer_kwargs=None, pre_dispatch='2*n_jobs', random_state=None,\n",
              "              refit=True, return_train_score=False, scoring='f1',\n",
              "              search_spaces={'logReg__C': Real(low=1e-05, high=100000.0, prior='log-uniform', transform='identity'),\n",
              "                             'logReg__l1_ratio': Real(low=0, high=1, prior='uniform', transform='identity')},\n",
              "              verbose=0)"
            ]
          },
          "metadata": {},
          "execution_count": 11
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "w4IShVevKCOM",
        "outputId": "bacd2039-ef6e-4432-807d-0118f2aa1a2c"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "source": [
        "# Optimisation du seuil de décision\n",
        "\n",
        "# y_true et prédictions\n",
        "y_true = train_df.TARGET_5Yrs\n",
        "y_pred = opt_logReg.predict_proba(train_df.drop(columns=[\"TARGET_5Yrs\"]))\n",
        "\n",
        "threshold = max_scores(\n",
        "    y_true,\n",
        "    [x[1] for x in y_pred],\n",
        "    min_range=10,\n",
        "    max_range=90\n",
        ")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Threshold = 0.33\n",
            "\n",
            "Scores:\n",
            "\n",
            "F1 score         = 0.7873\n",
            "Precision score  = 0.6976\n",
            "Recall score     = 0.9036\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0zDC6S0oizKP",
        "outputId": "395d090e-1703-44cf-9050-1879b951c7b2"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "source": [
        "# Optimisation du seuil de décision\n",
        "\n",
        "# y_true et prédictions\n",
        "y_true = train_df.TARGET_5Yrs\n",
        "y_pred = opt_logReg.predict_proba(train_df.drop(columns=[\"TARGET_5Yrs\"]))\n",
        "\n",
        "threshold = max_scores_fb(\n",
        "    y_true,\n",
        "    [x[1] for x in y_pred],\n",
        "    min_range=10,\n",
        "    max_range=90\n",
        ")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Threshold = 0.51\n",
            "\n",
            "Scores:\n",
            "\n",
            "fbeta_score = 0.7920\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qx-LQgiaeX3i",
        "outputId": "192dee9e-f9a7-41e3-d0c8-05e3e0c4c890"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "source": [
        "# prédictions jeu de test\n",
        "y_true = test_df.TARGET_5Yrs\n",
        "start = time.time()\n",
        "y_pred = opt_logReg.predict_proba(test_df.drop(columns=[\"TARGET_5Yrs\"]))\n",
        "stop = time.time()\n",
        "\n",
        "threshold = 0.51\n",
        "y_pred = [x[1] for x in y_pred]\n",
        "y_pred_threshold = [1 if x >= threshold else 0 for x in y_pred]\n",
        "\n",
        "print(f\"Recall    = {recall_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"F1 score  = {f1_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"Precision = {precision_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"Temps de prédictions = {stop - start}\")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Recall    = 0.5975\n",
            "F1 score  = 0.6597\n",
            "Precision = 0.7364\n",
            "Temps de prédictions = 0.002293825149536133\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2AF_SobLgsOH",
        "outputId": "e48b1dc6-0922-47b2-d53f-0998184cee5a"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "**SVM**"
      ],
      "metadata": {
        "id": "47nG9NjLTL5-"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# On cree un pipeline de traitement integrant la preparation et le modele Lasso:\n",
        "modele_svc = Pipeline([\n",
        "  (\"stdScaler\", StandardScaler()),\n",
        "  (\"svc\", SVC(\n",
        "      class_weight = class_weights_function(train_df.TARGET_5Yrs),\n",
        "      random_state = CONFIG[\"seed\"]\n",
        "  ))\n",
        "])\n",
        "\n",
        "# On definit les hyperparametres a faire varier:\n",
        "param_grid = {\n",
        "    \"svc__C\": Real(1e-5, 1e+5, prior=\"log-uniform\")\n",
        "}\n",
        "\n",
        "# On cree le modele utilisant une recherche bayesienne\n",
        "opt_svc = BayesSearchCV(\n",
        "    modele_svc,\n",
        "    param_grid,\n",
        "    scoring=\"f1\",\n",
        "    cv=CONFIG[\"nbr_folds\"],\n",
        "    n_jobs=-1,\n",
        "    n_iter=CONFIG[\"n_iter\"]\n",
        ")\n",
        "\n",
        "# On entraine le modele\n",
        "opt_svc.fit(\n",
        "  train_df.drop(columns=[\"TARGET_5Yrs\"]),\n",
        "  train_df.TARGET_5Yrs\n",
        ")"
      ],
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "BayesSearchCV(cv=5, error_score='raise',\n",
              "              estimator=Pipeline(memory=None,\n",
              "                                 steps=[('stdScaler',\n",
              "                                         StandardScaler(copy=True,\n",
              "                                                        with_mean=True,\n",
              "                                                        with_std=True)),\n",
              "                                        ('svc',\n",
              "                                         SVC(C=1.0, break_ties=False,\n",
              "                                             cache_size=200,\n",
              "                                             class_weight={0: 1.3220779220779222,\n",
              "                                                           1: 0.8041074249605055},\n",
              "                                             coef0=0.0,\n",
              "                                             decision_function_shape='ovr',\n",
              "                                             degree=3, gamma='scale',\n",
              "                                             kernel='rbf', max_iter=-1,\n",
              "                                             probabilit...se, random_state=0,\n",
              "                                             shrinking=True, tol=0.001,\n",
              "                                             verbose=False))],\n",
              "                                 verbose=False),\n",
              "              fit_params=None, iid=True, n_iter=66, n_jobs=-1, n_points=1,\n",
              "              optimizer_kwargs=None, pre_dispatch='2*n_jobs', random_state=None,\n",
              "              refit=True, return_train_score=False, scoring='f1',\n",
              "              search_spaces={'svc__C': Real(low=1e-05, high=100000.0, prior='log-uniform', transform='identity')},\n",
              "              verbose=0)"
            ]
          },
          "metadata": {},
          "execution_count": 20
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "aOcj-O-83k26",
        "outputId": "092c1419-ddc9-4487-ed80-515a0dd44177"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# Optimisation du seuil de décision\n",
        "\n",
        "# y_true et prédictions\n",
        "y_true = train_df.TARGET_5Yrs\n",
        "y_pred = opt_svc.decision_function(train_df.drop(columns=[\"TARGET_5Yrs\"]))\n",
        "\n",
        "threshold = max_scores(\n",
        "    y_true,\n",
        "    y_pred,\n",
        "    min_range=10,\n",
        "    max_range=90\n",
        ")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Threshold = 0.1\n",
            "\n",
            "Scores:\n",
            "\n",
            "F1 score         = 0.7309\n",
            "Precision score  = 0.8112\n",
            "Recall score     = 0.6651\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0zLZNqcPgWYi",
        "outputId": "16d4110a-50cf-4459-85b2-10460795c164"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# prédictions jeu de test\n",
        "y_true = test_df.TARGET_5Yrs\n",
        "start = time.time()\n",
        "y_pred = opt_svc.decision_function(test_df.drop(columns=[\"TARGET_5Yrs\"]))\n",
        "stop = time.time()\n",
        "\n",
        "threshold = 0.1\n",
        "y_pred_threshold = [1 if x >= threshold else 0 for x in y_pred]\n",
        "\n",
        "print(f\"Recall    = {recall_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"F1 score  = {f1_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"Precision = {precision_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"Temps de prédictions = {stop - start}\")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Recall    = 0.5660\n",
            "F1 score  = 0.6406\n",
            "Precision = 0.7377\n",
            "Temps de prédictions = 0.014455080032348633\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ypYIt3GEhnPd",
        "outputId": "0c1d69f1-5553-451f-c985-fc740b42dc28"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "**Random Forest et Gradient Boosting**\n",
        "\n",
        "Il y a un problème lors de l'utilisation de skopt-Bayes_search_cv et un modèle ensembliste.\n",
        "Une solution a été proposée [ici](https://github.com/scikit-optimize/scikit-optimize/issues/978#issue-775632361) par l'utilisateur *AyrtonB*.\n",
        "\n",
        "Cependant, cette modélisation est très couteuse en temps.\n",
        "Je fais alors le choix de ne pas entraîner de modèles ensemblistes et passe directement à une modélisation utilisant de l'autoML."
      ],
      "metadata": {
        "id": "kRxBIl7PT01L"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "**AutoML - AutoGluon**\n",
        "\n",
        "Lors de l'installation d'AutGluon, je suspecte qu'il y ait une modification de certaines librairies, renant impossible l'utilisation de scikit-optimize avec les modèles de sklearn entraînés précédemment.\n",
        "\n",
        "De ce fait, pour éviter tout problème, je préfère installer et importer AutoGluon seulement à partir d'ici dans le NoteBook."
      ],
      "metadata": {
        "id": "NpeO8NlbT9Rn"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# AutoML (AutoGluon)\n",
        "!pip install -U pip\n",
        "!pip install -U setuptools wheel\n",
        "!pip install -U \"mxnet<2.0.0\"\n",
        "!pip install autogluon.tabular\n",
        "\n",
        "!pip install catboost"
      ],
      "outputs": [],
      "metadata": {
        "id": "KtJKfM3MIUgL"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "from autogluon.tabular import TabularPredictor"
      ],
      "outputs": [],
      "metadata": {
        "id": "H-ILWVgqjJDk"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# Entraînement du modèle AutoGluon\n",
        "\n",
        "predictor = TabularPredictor(\n",
        "    label=\"TARGET_5Yrs\",\n",
        "    problem_type=\"binary\",\n",
        "    eval_metric=\"f1\",\n",
        "    sample_weight=\"auto_weight\"\n",
        ").fit(\n",
        "   train_df,\n",
        "   time_limit=900,\n",
        "   presets='best_quality'\n",
        ")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "No path specified. Models will be saved in: \"AutogluonModels/ag-20210903_130843/\"\n",
            "Presets specified: ['best_quality']\n",
            "Beginning AutoGluon training ... Time limit = 900s\n",
            "AutoGluon will save models to \"AutogluonModels/ag-20210903_130843/\"\n",
            "AutoGluon Version:  0.3.1\n",
            "Train Data Rows:    1018\n",
            "Train Data Columns: 11\n",
            "Preprocessing data ...\n",
            "NumExpr defaulting to 2 threads.\n",
            "Selected class <--> label mapping:  class 1 = 1, class 0 = 0\n",
            "Using Feature Generators to preprocess the data ...\n",
            "Fitting AutoMLPipelineFeatureGenerator...\n",
            "\tAvailable Memory:                    12692.97 MB\n",
            "\tTrain Data (Original)  Memory Usage: 0.09 MB (0.0% 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', []) : 10 | ['PTS', 'FG%', '3P%', 'FT%', 'OREB', ...]\n",
            "\t\t('int', [])   :  1 | ['GP']\n",
            "\tTypes of features in processed data (raw dtype, special dtypes):\n",
            "\t\t('float', []) : 10 | ['PTS', 'FG%', '3P%', 'FT%', 'OREB', ...]\n",
            "\t\t('int', [])   :  1 | ['GP']\n",
            "\t0.1s = Fit runtime\n",
            "\t11 features in original data used to generate 11 features in processed data.\n",
            "\tTrain Data (Processed) Memory Usage: 0.09 MB (0.0% of available memory)\n",
            "Data preprocessing and feature engineering runtime = 0.11s ...\n",
            "AutoGluon will gauge predictive performance using evaluation metric: 'f1'\n",
            "\tTo change this, specify the eval_metric argument of fit()\n",
            "AutoGluon will fit 2 stack levels (L1 to L2) ...\n",
            "Fitting 13 L1 models ...\n",
            "Fitting model: KNeighborsUnif_BAG_L1 ... Training model for up to 599.77s of the 899.88s of remaining time.\n",
            "\t0.7586\t = Validation score   (f1)\n",
            "\t0.01s\t = Training   runtime\n",
            "\t0.1s\t = Validation runtime\n",
            "Fitting model: KNeighborsDist_BAG_L1 ... Training model for up to 599.63s of the 899.73s of remaining time.\n",
            "\t0.7513\t = Validation score   (f1)\n",
            "\t0.01s\t = Training   runtime\n",
            "\t0.1s\t = Validation runtime\n",
            "Fitting model: LightGBMXT_BAG_L1 ... Training model for up to 599.49s of the 899.59s of remaining time.\n"
          ]
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-11-3a27b4b16925>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      9\u001b[0m    \u001b[0mtrain_df\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     10\u001b[0m    \u001b[0mtime_limit\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m900\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 11\u001b[0;31m    \u001b[0mpresets\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'best_quality'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     12\u001b[0m )\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/core/utils/decorators.py\u001b[0m in \u001b[0;36m_call\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m     27\u001b[0m         \u001b[0;32mdef\u001b[0m \u001b[0m_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     28\u001b[0m             \u001b[0mgargs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgkwargs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mg\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mother_args\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 29\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mgargs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mgkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     30\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0m_call\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     31\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0m_unpack_inner\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/predictor/predictor.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, train_data, tuning_data, time_limit, presets, hyperparameters, feature_metadata, **kwargs)\u001b[0m\n\u001b[1;32m    745\u001b[0m         self._learner.fit(X=train_data, X_val=tuning_data, X_unlabeled=unlabeled_data,\n\u001b[1;32m    746\u001b[0m                           \u001b[0mholdout_frac\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mholdout_frac\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_bag_folds\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mnum_bag_folds\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_bag_sets\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mnum_bag_sets\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_stack_levels\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mnum_stack_levels\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 747\u001b[0;31m                           hyperparameters=hyperparameters, core_kwargs=core_kwargs, time_limit=time_limit, verbosity=verbosity, use_bag_holdout=use_bag_holdout)\n\u001b[0m\u001b[1;32m    748\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_set_post_fit_vars\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    749\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/learner/abstract_learner.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, X, X_val, **kwargs)\u001b[0m\n\u001b[1;32m    147\u001b[0m             \u001b[0;32mraise\u001b[0m \u001b[0mAssertionError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Learner is already fit.'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    148\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_validate_fit_input\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mX_val\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 149\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mX_val\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    150\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    151\u001b[0m     def _fit(self, X: DataFrame, X_val: DataFrame = None, scheduler_options=None, hyperparameter_tune=False,\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/learner/default_learner.py\u001b[0m in \u001b[0;36m_fit\u001b[0;34m(self, X, X_val, X_unlabeled, holdout_frac, num_bag_folds, num_bag_sets, time_limit, verbosity, **trainer_fit_kwargs)\u001b[0m\n\u001b[1;32m     97\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     98\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 99\u001b[0;31m         \u001b[0mtrainer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mX_val\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_val\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mX_unlabeled\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX_unlabeled\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mholdout_frac\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mholdout_frac\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtime_limit\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtime_limit_trainer\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgroups\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mgroups\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mtrainer_fit_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    100\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave_trainer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrainer\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtrainer\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    101\u001b[0m         \u001b[0mtime_end\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtime\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtime\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/auto_trainer.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, X, y, hyperparameters, X_val, y_val, X_unlabeled, feature_prune, holdout_frac, num_stack_levels, core_kwargs, time_limit, use_bag_holdout, groups, **kwargs)\u001b[0m\n\u001b[1;32m     72\u001b[0m         self._train_multi_and_ensemble(X, y, X_val, y_val, X_unlabeled=X_unlabeled, hyperparameters=hyperparameters,\n\u001b[1;32m     73\u001b[0m                                        \u001b[0mfeature_prune\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfeature_prune\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 74\u001b[0;31m                                        num_stack_levels=num_stack_levels, time_limit=time_limit, core_kwargs=core_kwargs, groups=groups)\n\u001b[0m\u001b[1;32m     75\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     76\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mconstruct_model_templates_distillation\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhyperparameters\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36m_train_multi_and_ensemble\u001b[0;34m(self, X, y, X_val, y_val, hyperparameters, X_unlabeled, num_stack_levels, time_limit, groups, **kwargs)\u001b[0m\n\u001b[1;32m   1378\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_num_cols_train\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlist\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1379\u001b[0m         model_names_fit = self.train_multi_levels(X, y, hyperparameters=hyperparameters, X_val=X_val, y_val=y_val,\n\u001b[0;32m-> 1380\u001b[0;31m                                                   X_unlabeled=X_unlabeled, level_start=1, level_end=num_stack_levels+1, time_limit=time_limit, **kwargs)\n\u001b[0m\u001b[1;32m   1381\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_model_names\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1382\u001b[0m             \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'AutoGluon did not successfully train any models'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36mtrain_multi_levels\u001b[0;34m(self, X, y, hyperparameters, X_val, y_val, X_unlabeled, base_model_names, feature_prune, core_kwargs, aux_kwargs, level_start, level_end, time_limit, name_suffix, relative_stack, level_time_modifier)\u001b[0m\n\u001b[1;32m    288\u001b[0m                 \u001b[0mmodels\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mhyperparameters\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlevel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mlevel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbase_model_names\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mbase_model_names\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    289\u001b[0m                 \u001b[0mfeature_prune\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfeature_prune\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 290\u001b[0;31m                 \u001b[0mcore_kwargs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcore_kwargs_level\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maux_kwargs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0maux_kwargs_level\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname_suffix\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mname_suffix\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    291\u001b[0m             )\n\u001b[1;32m    292\u001b[0m             \u001b[0mmodel_names_fit\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0mbase_model_names\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0maux_models\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36mstack_new_level\u001b[0;34m(self, X, y, models, X_val, y_val, X_unlabeled, level, base_model_names, feature_prune, core_kwargs, aux_kwargs, name_suffix)\u001b[0m\n\u001b[1;32m    314\u001b[0m             \u001b[0maux_kwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'name_suffix'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0maux_kwargs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'name_suffix'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m''\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mname_suffix\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    315\u001b[0m         core_models = self.stack_new_level_core(X=X, y=y, X_val=X_val, y_val=y_val, X_unlabeled=X_unlabeled, models=models,\n\u001b[0;32m--> 316\u001b[0;31m                                                 level=level, base_model_names=base_model_names, feature_prune=feature_prune, **core_kwargs)\n\u001b[0m\u001b[1;32m    317\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    318\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mX_val\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36mstack_new_level_core\u001b[0;34m(self, X, y, models, X_val, y_val, X_unlabeled, level, base_model_names, stack_name, ag_args, ag_args_fit, ag_args_ensemble, excluded_model_types, ensemble_type, name_suffix, get_models_func, refit_full, **kwargs)\u001b[0m\n\u001b[1;32m    394\u001b[0m         \u001b[0;31m# FIXME: TODO: v0.1 X_unlabeled isn't cached so it won't be available during refit_full or fit_extra.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    395\u001b[0m         return self._train_multi(X=X_init, y=y, X_val=X_val, y_val=y_val, X_unlabeled=X_unlabeled,\n\u001b[0;32m--> 396\u001b[0;31m                                  models=models, level=level, stack_name=stack_name, fit_kwargs=fit_kwargs, **kwargs)\n\u001b[0m\u001b[1;32m    397\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    398\u001b[0m     \u001b[0;31m# TODO: Consider making level be auto-determined based off of max(base_model_levels)+1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36m_train_multi\u001b[0;34m(self, X, y, models, hyperparameter_tune_kwargs, feature_prune, k_fold, n_repeats, n_repeat_start, time_limit, **kwargs)\u001b[0m\n\u001b[1;32m   1350\u001b[0m             \u001b[0mtime_start\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtime\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtime\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1351\u001b[0m             model_names_trained = self._train_multi_initial(X=X, y=y, models=models, k_fold=k_fold, n_repeats=n_repeats_initial, hyperparameter_tune_kwargs=hyperparameter_tune_kwargs, feature_prune=feature_prune,\n\u001b[0;32m-> 1352\u001b[0;31m                                                             time_limit=time_limit, **kwargs)\n\u001b[0m\u001b[1;32m   1353\u001b[0m             \u001b[0mn_repeat_start\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mn_repeats_initial\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1354\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mtime_limit\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36m_train_multi_initial\u001b[0;34m(self, X, y, models, k_fold, n_repeats, hyperparameter_tune_kwargs, feature_prune, time_limit, **kwargs)\u001b[0m\n\u001b[1;32m   1278\u001b[0m                     \u001b[0mtime_limit\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtime_limit\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mtime\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtime\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mtime_start\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1279\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1280\u001b[0;31m             \u001b[0mmodels\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_train_multi_fold\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodels\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmodels\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhyperparameter_tune_kwargs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeature_prune\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mk_fold_start\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mk_fold_start\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mk_fold_end\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mk_fold\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mn_repeats\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mn_repeats\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mn_repeat_start\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtime_limit\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtime_limit\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mfit_args\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1281\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1282\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mmodels\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36m_train_multi_fold\u001b[0;34m(self, X, y, models, time_limit, time_split, time_ratio, hyperparameter_tune_kwargs, **kwargs)\u001b[0m\n\u001b[1;32m   1320\u001b[0m                     \u001b[0mtime_left\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtime_limit\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mtime_start_model\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mtime_start\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1321\u001b[0m             model_name_trained_lst = self._train_single_full(X, y, model, time_limit=time_left,\n\u001b[0;32m-> 1322\u001b[0;31m                                                              hyperparameter_tune_kwargs=hyperparameter_tune_kwargs_model, **kwargs)\n\u001b[0m\u001b[1;32m   1323\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1324\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlow_memory\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36m_train_single_full\u001b[0;34m(self, X, y, model, X_unlabeled, X_val, y_val, feature_prune, hyperparameter_tune_kwargs, stack_name, k_fold, k_fold_start, k_fold_end, n_repeats, n_repeat_start, level, time_limit, fit_kwargs, **kwargs)\u001b[0m\n\u001b[1;32m   1179\u001b[0m                     \u001b[0mcompute_base_preds\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1180\u001b[0m                 ))\n\u001b[0;32m-> 1181\u001b[0;31m             \u001b[0mmodel_names_trained\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_train_and_save\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mX_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mX_unlabeled\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX_unlabeled\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstack_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mstack_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlevel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mlevel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mmodel_fit_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1182\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1183\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mmodel_names_trained\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36m_train_and_save\u001b[0;34m(self, X, y, model, X_val, y_val, stack_name, level, **model_fit_kwargs)\u001b[0m\n\u001b[1;32m    960\u001b[0m                 \u001b[0mfit_log_message\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0;34mf' Training model for up to {round(time_limit, 2)}s of the {round(time_left_total, 2)}s of remaining time.'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    961\u001b[0m             \u001b[0mlogger\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlog\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m20\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfit_log_message\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 962\u001b[0;31m             \u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_train_single\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mX_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mmodel_fit_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    963\u001b[0m             \u001b[0mfit_end_time\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtime\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtime\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    964\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mweight_evaluation\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/trainer/abstract_trainer.py\u001b[0m in \u001b[0;36m_train_single\u001b[0;34m(self, X, y, model, X_val, y_val, **model_fit_kwargs)\u001b[0m\n\u001b[1;32m    932\u001b[0m         \u001b[0mReturns\u001b[0m \u001b[0mtrained\u001b[0m \u001b[0mmodel\u001b[0m \u001b[0mobject\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    933\u001b[0m         \"\"\"\n\u001b[0;32m--> 934\u001b[0;31m         \u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mX_val\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_val\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mmodel_fit_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    935\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    936\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/core/models/abstract/abstract_model.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, **kwargs)\u001b[0m\n\u001b[1;32m    520\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;34m'time_limit'\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mkwargs\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'time_limit'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'time_limit'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    521\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_register_fit_metadata\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 522\u001b[0;31m             \u001b[0mout\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    523\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mout\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    524\u001b[0m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/core/models/ensemble/stacker_ensemble_model.py\u001b[0m in \u001b[0;36m_fit\u001b[0;34m(self, X, y, compute_base_preds, time_limit, **kwargs)\u001b[0m\n\u001b[1;32m    151\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mtime_limit\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    152\u001b[0m             \u001b[0mtime_limit\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtime_limit\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mtime\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtime\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mstart_time\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 153\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtime_limit\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtime_limit\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    154\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    155\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mset_contexts\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpath_context\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/core/models/ensemble/bagged_ensemble_model.py\u001b[0m in \u001b[0;36m_fit\u001b[0;34m(self, X, y, X_val, y_val, k_fold, k_fold_start, k_fold_end, n_repeats, n_repeat_start, groups, **kwargs)\u001b[0m\n\u001b[1;32m    187\u001b[0m                     \u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'time_limit'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'time_limit'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mfolds_to_fit\u001b[0m \u001b[0;34m/\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mfolds_to_fit\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;36m1.2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    188\u001b[0m             self._fit_folds(X=X, y=y, model_base=model_base, k_fold=k_fold, k_fold_start=k_fold_start, k_fold_end=k_fold_end,\n\u001b[0;32m--> 189\u001b[0;31m                             n_repeats=n_repeats, n_repeat_start=n_repeat_start, save_folds=save_bag_folds, groups=groups, **kwargs)\n\u001b[0m\u001b[1;32m    190\u001b[0m             \u001b[0;31m# FIXME: Don't save folds except for refit\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    191\u001b[0m             \u001b[0;31m# FIXME: Cleanup self\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/core/models/ensemble/bagged_ensemble_model.py\u001b[0m in \u001b[0;36m_fit_folds\u001b[0;34m(self, X, y, model_base, k_fold, k_fold_start, k_fold_end, n_repeats, n_repeat_start, time_limit, sample_weight, save_folds, groups, **kwargs)\u001b[0m\n\u001b[1;32m    386\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mfold_end_n_repeat\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0mfold_end\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mk_fold\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0mk_fold_end\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    387\u001b[0m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_k_per_n_repeat\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mk_fold\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 388\u001b[0;31m         \u001b[0mfold_fitting_strategy\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mafter_all_folds_scheduled\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    389\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodels\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0mmodels\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    390\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/core/models/ensemble/fold_fitting_strategy.py\u001b[0m in \u001b[0;36mafter_all_folds_scheduled\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m     57\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mafter_all_folds_scheduled\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     58\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mjob\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mjobs\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 59\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_fit_fold_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mjob\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     60\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     61\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_fit_fold_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel_base\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfold_ctx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/core/models/ensemble/fold_fitting_strategy.py\u001b[0m in \u001b[0;36m_fit_fold_model\u001b[0;34m(self, model_base, fold_ctx, kwargs)\u001b[0m\n\u001b[1;32m     62\u001b[0m         \u001b[0mtime_start_fold\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtime\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtime\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     63\u001b[0m         \u001b[0mtime_limit_fold\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_get_fold_time_limit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfold_ctx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 64\u001b[0;31m         \u001b[0mfold_model\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel_base\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtime_start_fold\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtime_limit_fold\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfold_ctx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     65\u001b[0m         \u001b[0mfold_model\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpred_proba\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_predict_oof\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfold_model\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfold_ctx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     66\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_update_bagged_ensemble\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfold_model\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpred_proba\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfold_ctx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/core/models/ensemble/fold_fitting_strategy.py\u001b[0m in \u001b[0;36m_fit\u001b[0;34m(self, model_base, time_start_fold, time_limit_fold, fold_ctx, kwargs)\u001b[0m\n\u001b[1;32m    109\u001b[0m             \u001b[0mkwargs_fold\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'sample_weight'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msample_weight\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mtrain_index\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    110\u001b[0m             \u001b[0mkwargs_fold\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'sample_weight_val'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msample_weight\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mval_index\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 111\u001b[0;31m         \u001b[0mfold_model\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX_fold\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my_fold\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mX_val\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX_val_fold\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_val\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my_val_fold\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtime_limit\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtime_limit_fold\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs_fold\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    112\u001b[0m         \u001b[0mfold_model\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit_time\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtime\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtime\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mtime_start_fold\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    113\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mfold_model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/core/models/abstract/abstract_model.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, **kwargs)\u001b[0m\n\u001b[1;32m    520\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;34m'time_limit'\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mkwargs\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'time_limit'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'time_limit'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    521\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_register_fit_metadata\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 522\u001b[0;31m             \u001b[0mout\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    523\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mout\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    524\u001b[0m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/autogluon/tabular/models/lgb/lgb_model.py\u001b[0m in \u001b[0;36m_fit\u001b[0;34m(self, X, y, X_val, y_val, dataset_train, dataset_val, time_limit, num_gpus, sample_weight, sample_weight_val, verbosity, **kwargs)\u001b[0m\n\u001b[1;32m    180\u001b[0m             \u001b[0mwarnings\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfilterwarnings\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'ignore'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmessage\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'categorical_column in param dict is overridden.'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    181\u001b[0m             \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 182\u001b[0;31m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlgb\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mtrain_params\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    183\u001b[0m             \u001b[0;32mexcept\u001b[0m \u001b[0mLightGBMError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    184\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0mtrain_params\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'params'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'device'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'cpu'\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0;34m'gpu'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/lightgbm/engine.py\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(params, train_set, num_boost_round, valid_sets, valid_names, fobj, feval, init_model, feature_name, categorical_feature, early_stopping_rounds, evals_result, verbose_eval, learning_rates, keep_training_booster, callbacks)\u001b[0m\n\u001b[1;32m    223\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mis_valid_contain_train\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    224\u001b[0m                 \u001b[0mevaluation_result_list\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mextend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbooster\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0meval_train\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfeval\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 225\u001b[0;31m             \u001b[0mevaluation_result_list\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mextend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbooster\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0meval_valid\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfeval\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    226\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    227\u001b[0m             \u001b[0;32mfor\u001b[0m \u001b[0mcb\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mcallbacks_after_iter\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/lightgbm/basic.py\u001b[0m in \u001b[0;36meval_valid\u001b[0;34m(self, feval)\u001b[0m\n\u001b[1;32m   1975\u001b[0m             \u001b[0mList\u001b[0m \u001b[0;32mwith\u001b[0m \u001b[0mevaluation\u001b[0m \u001b[0mresults\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1976\u001b[0m         \"\"\"\n\u001b[0;32m-> 1977\u001b[0;31m         return [item for i in range_(1, self.__num_dataset)\n\u001b[0m\u001b[1;32m   1978\u001b[0m                 for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]\n\u001b[1;32m   1979\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/lightgbm/basic.py\u001b[0m in \u001b[0;36m<listcomp>\u001b[0;34m(.0)\u001b[0m\n\u001b[1;32m   1976\u001b[0m         \"\"\"\n\u001b[1;32m   1977\u001b[0m         return [item for i in range_(1, self.__num_dataset)\n\u001b[0;32m-> 1978\u001b[0;31m                 for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]\n\u001b[0m\u001b[1;32m   1979\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1980\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0msave_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfilename\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_iteration\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstart_iteration\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/lightgbm/basic.py\u001b[0m in \u001b[0;36m__inner_eval\u001b[0;34m(self, data_name, data_idx, feval)\u001b[0m\n\u001b[1;32m   2362\u001b[0m                 \u001b[0mctypes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mc_int\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata_idx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2363\u001b[0m                 \u001b[0mctypes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbyref\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtmp_out_len\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2364\u001b[0;31m                 result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))\n\u001b[0m\u001b[1;32m   2365\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mtmp_out_len\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalue\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__num_inner_eval\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2366\u001b[0m                 \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Wrong length of eval results\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "2ToY4zZH330O",
        "outputId": "441ef829-503d-4ba7-b97e-d09118a0f74a"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# Évalutation des performances du modèle avant optimisation seuil de décision\n",
        "start = time.time()\n",
        "performance = predictor.evaluate(test_df)\n",
        "stop = time.time()\n",
        "\n",
        "print(f\"\\nTemps de prédictions = {stop - start}\")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Evaluation: f1 on test data: 0.7522388059701492\n",
            "Evaluations on test data:\n",
            "{\n",
            "    \"f1\": 0.7522388059701492,\n",
            "    \"accuracy\": 0.6745098039215687,\n",
            "    \"balanced_accuracy\": 0.635809748427673,\n",
            "    \"mcc\": 0.2845936721958216,\n",
            "    \"roc_auc\": 0.6864517819706499,\n",
            "    \"precision\": 0.7159090909090909,\n",
            "    \"recall\": 0.7924528301886793\n",
            "}\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "Temps de prédictions = 16.251795530319214\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3_xvn496FIjY",
        "outputId": "a1e70fce-7d1a-4bea-91d1-87bccd02c624"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# Optimisation du seuil de décision\n",
        "\n",
        "# y_true et prédictions\n",
        "y_true = train_df.TARGET_5Yrs\n",
        "y_pred = predictor.predict_proba(train_df.drop(columns=[\"TARGET_5Yrs\"]))[1]\n",
        "\n",
        "threshold = max_scores(\n",
        "    y_true,\n",
        "    y_pred,\n",
        "    min_range=10,\n",
        "    max_range=90\n",
        ")"
      ],
      "outputs": [],
      "metadata": {
        "id": "pynBGKYRisJu"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# prédictions jeu de test\n",
        "y_true = test_df.TARGET_5Yrs\n",
        "y_pred = predictor.predict_proba(test_df.drop(columns=[\"TARGET_5Yrs\"]))[1]\n",
        "\n",
        "threshold = 0.27\n",
        "y_pred_threshold = [1 if x >= threshold else 0 for x in y_pred]\n",
        "\n",
        "print(f\"Recall    = {recall_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"F1 score  = {f1_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"Precision = {precision_score(y_true, y_pred_threshold):.4f}\")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Recall    = 0.9874\n",
            "F1 score  = 0.7696\n",
            "Precision = 0.6305\n"
          ]
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hFbV4pY7iten",
        "outputId": "2f6c3275-4388-489d-e882-9c62af11bd4d"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Comparaison des différents modèles"
      ],
      "metadata": {
        "id": "H3ZgkiqoqNlc"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "source": [
        "list_model_name = [\"Baseline\", \"Log_reg\", \"SVM\", \"AutoML\"]\n",
        "list_test_f1 = [0.5761, 0.6597, 0.6406, 0.7696]\n",
        "list_test_precision = [0.5933, 0.7364, 0.7377, 0.6305]\n",
        "liste_temps_pred = [0.0022, 0.0023, 0.0145, 16.25]\n",
        "\n",
        "# Visualization\n",
        "plt.style.use(['default'])\n",
        "f, axes = plt.subplots(1, 3, figsize=(20, 6))\n",
        "\n",
        "# Test - F1 score\n",
        "axes[0].barh(\n",
        "    list_model_name,\n",
        "    list_test_f1,\n",
        "    color=plt.cm.Paired(np.arange(len(list_test_f1)))\n",
        ")\n",
        "axes[0].set_title('Test - F1 score')\n",
        "axes[0].set_yticklabels([])\n",
        "axes[0].invert_yaxis()\n",
        "\n",
        "# Test - Precision\n",
        "axes[1].barh(\n",
        "    list_model_name,\n",
        "    list_test_precision,\n",
        "    color=plt.cm.Paired(np.arange(len(list_test_precision)))\n",
        ")\n",
        "axes[1].set_title('Test - Precision')\n",
        "axes[1].set_yticklabels([])\n",
        "axes[1].invert_yaxis()\n",
        "\n",
        "# Temps pred jeu de test\n",
        "axes[2].barh(\n",
        "    list_model_name,\n",
        "    liste_temps_pred,\n",
        "    color=plt.cm.Paired(np.arange(len(liste_temps_pred)))\n",
        ")\n",
        "axes[2].set_title('Temps pred jeu de test')\n",
        "axes[2].set_yticklabels([])\n",
        "axes[2].invert_yaxis()"
      ],
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 2000x600 with 3 Axes>"
            ]
          },
          "metadata": {}
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "id": "2izvM6q5qMcD",
        "outputId": "dfd893d0-8e6a-4a7d-b842-811bbdd9fae3"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "La régression logistique est dépassée par AutoGluon pour la valeur du F1-score.\n",
        "\n",
        "Or, sa précision, ainsi que le temps de prédiction sont meilleurs la concernant.\n",
        "\n",
        "C'est le modèle que je choisis par la suite."
      ],
      "metadata": {
        "id": "KtJ96jBZJfOF"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Meilleur modèle\n",
        "\n",
        "Notre meilleur modèle est la régression logistique.\n",
        "\n",
        "Nous allons à présent optimiser le seuil de décision afin de maximiser le F0.5-score.\n",
        "\n",
        "Puis, nous afficherons la matrice de confusion produite par ce modèle, en utilisant le seuil optimal.\n",
        "\n",
        "Enfin, nous enregistrerons notre modèle à l'aide de \"dill\"."
      ],
      "metadata": {
        "id": "iPctT8No8tbD"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 25,
      "source": [
        "# Matrice de confusion\n",
        "# optimisation du seuil sur le jeu d'entrainement\n",
        "plt.style.use(['default'])\n",
        "\n",
        "y_true = train_df.TARGET_5Yrs\n",
        "y_pred = opt_logReg.predict_proba(train_df.drop(columns=[\"TARGET_5Yrs\"]))\n",
        "\n",
        "threshold = 0.51\n",
        "y_pred = [x[1] for x in y_pred]\n",
        "y_pred_threshold = [1 if x >= threshold else 0 for x in y_pred]\n",
        "\n",
        "ax = sns.heatmap(\n",
        "    confusion_matrix(y_true, y_pred_threshold),\n",
        "    annot=True,\n",
        "    fmt=\"d\"\n",
        ")\n",
        "\n",
        "ax.set_xlabel(\"Predicted labels\");\n",
        "ax.set_ylabel(\"True labels\");\n",
        "ax.set_title(\"Matrice de confusion - Contrat NBA\");\n",
        "ax.xaxis.set_ticklabels([\"carreer length < 5\", \"carreer length >= 5\"]);\n",
        "ax.yaxis.set_ticklabels([\"carreer length < 5\", \"carreer length >= 5\"]);\n",
        "\n",
        "print(f\"Recall    = {recall_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"F{0.5}-score = {fbeta_score(y_true, y_pred_threshold, beta=0.5):.4f}\")\n",
        "print(f\"Precision = {precision_score(y_true, y_pred_threshold):.4f}\\n\")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Recall    = 0.6919\n",
            "F0.5-score = 0.7920\n",
            "Precision = 0.8218\n",
            "\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 2 Axes>"
            ]
          },
          "metadata": {}
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 540
        },
        "id": "Ykf0EXe59Jdi",
        "outputId": "a7e556b1-db85-4ae6-aa25-a79e206815d2"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 26,
      "source": [
        "# Matrice de confusion\n",
        "# Resulats sur le jeu de test\n",
        "\n",
        "plt.style.use(['default'])\n",
        "\n",
        "y_true = test_df.TARGET_5Yrs\n",
        "y_pred = opt_logReg.predict_proba(test_df.drop(columns=[\"TARGET_5Yrs\"]))\n",
        "\n",
        "threshold = 0.51\n",
        "y_pred = [x[1] for x in y_pred]\n",
        "y_pred_threshold = [1 if x >= threshold else 0 for x in y_pred]\n",
        "\n",
        "ax = sns.heatmap(\n",
        "    confusion_matrix(y_true, y_pred_threshold),\n",
        "    annot=True,\n",
        "    fmt=\"d\"\n",
        ")\n",
        "\n",
        "ax.set_xlabel(\"Predicted labels\");\n",
        "ax.set_ylabel(\"True labels\");\n",
        "ax.set_title(\"Matrice de confusion - Contrat NBA\");\n",
        "ax.xaxis.set_ticklabels([\"carreer length < 5\", \"carreer length >= 5\"]);\n",
        "ax.yaxis.set_ticklabels([\"carreer length < 5\", \"carreer length >= 5\"]);\n",
        "\n",
        "print(f\"Recall    = {recall_score(y_true, y_pred_threshold):.4f}\")\n",
        "print(f\"F{0.5}-score = {fbeta_score(y_true, y_pred_threshold, beta=0.5):.4f}\")\n",
        "print(f\"Precision = {precision_score(y_true, y_pred_threshold):.4f}\\n\")"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Recall    = 0.5975\n",
            "F0.5-score = 0.7037\n",
            "Precision = 0.7364\n",
            "\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 2 Axes>"
            ]
          },
          "metadata": {}
        }
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 540
        },
        "id": "5hZrfRikiRnv",
        "outputId": "af8a0608-7e2e-42a5-b1a5-1936f9178617"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# Enregistrement du modèle\n",
        "path = \"/content/drive/MyDrive/Gaetan_Travail/ML/\"\n",
        "\n",
        "with open(path + \"model_logReg.pk\", \"wb\") as file:\n",
        "\tpickle.dump(opt_logReg, file)"
      ],
      "outputs": [],
      "metadata": {
        "id": "p7tdn-APaKHP"
      }
    }
  ]
}