stats548557_gbm_overfitting_ntrees.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "stats548557-gbm_overfitting_ntrees.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyOVREUMPAfHIUzA9v8UVzCy",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/bmreiniger/datascience.stackexchange/blob/master/stats548557_gbm_overfitting_ntrees.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Yww7QEdKh2RP"
      },
      "source": [
        "from sklearn.ensemble import GradientBoostingRegressor\n",
        "from sklearn.datasets import fetch_california_housing\n",
        "from sklearn.model_selection import train_test_split"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0xrZVDMpiAfx",
        "outputId": "feb28ce5-1c1c-4ab2-9015-e0e3baeb4c55"
      },
      "source": [
        "X, y = fetch_california_housing(return_X_y=True)"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Downloading Cal. housing from https://ndownloader.figshare.com/files/5976036 to /root/scikit_learn_data\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_AW8IzEgiGlM"
      },
      "source": [
        "X_train, X_test, y_train, y_test = train_test_split(X, y)"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Q7UjHsa1iOGa",
        "outputId": "8c2ae964-2263-4ea8-a752-3c2c10f29a51"
      },
      "source": [
        "gbm = GradientBoostingRegressor(max_leaf_nodes=6, learning_rate=0.1, n_estimators=10000, loss='huber')\n",
        "gbm.fit(X_train, y_train)"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GradientBoostingRegressor(alpha=0.9, ccp_alpha=0.0, criterion='friedman_mse',\n",
              "                          init=None, learning_rate=0.1, loss='huber',\n",
              "                          max_depth=3, max_features=None, max_leaf_nodes=6,\n",
              "                          min_impurity_decrease=0.0, min_impurity_split=None,\n",
              "                          min_samples_leaf=1, min_samples_split=2,\n",
              "                          min_weight_fraction_leaf=0.0, n_estimators=10000,\n",
              "                          n_iter_no_change=None, presort='deprecated',\n",
              "                          random_state=None, subsample=1.0, tol=0.0001,\n",
              "                          validation_fraction=0.1, verbose=0, warm_start=False)"
            ]
          },
          "metadata": {},
          "execution_count": 30
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 441
        },
        "id": "BVGsiy0Ziaka",
        "outputId": "12a7e885-abe3-4c95-d4eb-e4ed672f2cad"
      },
      "source": [
        "import numpy as np\n",
        "from matplotlib import pyplot as plt\n",
        "from sklearn.metrics import mean_absolute_error\n",
        "\n",
        "model = gbm\n",
        "n_iter = model.n_estimators_\n",
        "\n",
        "train_score = np.zeros((n_iter,), dtype=np.float64)\n",
        "test_score = np.zeros((n_iter,), dtype=np.float64)\n",
        "for i, y_pred in enumerate(model.staged_predict(X_train)):\n",
        "    train_score[i] = mean_absolute_error(y_train, y_pred)\n",
        "for i, y_pred in enumerate(model.staged_predict(X_test)):\n",
        "    test_score[i] = mean_absolute_error(y_test, y_pred)\n",
        "\n",
        "fig = plt.figure(figsize=(6, 6))\n",
        "plt.subplot(1, 1, 1)\n",
        "plt.title('Deviance')\n",
        "plt.plot(np.arange(n_iter) + 1, train_score, 'b-',\n",
        "         label='Training Set Deviance')\n",
        "plt.plot(np.arange(n_iter) + 1, test_score, 'r-',\n",
        "         label='Test Set Deviance')\n",
        "plt.legend(loc='upper right')\n",
        "plt.xlabel('Boosting Iterations')\n",
        "plt.ylabel('Deviance')\n",
        "plt.ylim(0, 0.4)\n",
        "fig.tight_layout()\n",
        "plt.show()\n"
      ],
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 441
        },
        "id": "yb7X1Znuqf1Q",
        "outputId": "db718b93-3cdc-4e6e-e9e4-0ba2c95d8dd5"
      },
      "source": [
        "gbm_noshrink = GradientBoostingRegressor(max_leaf_nodes=6, learning_rate=1.0, n_estimators=500)\n",
        "gbm_noshrink.fit(X_train, y_train)\n",
        "\n",
        "model = gbm_noshrink\n",
        "\n",
        "n_iter = model.n_estimators_\n",
        "\n",
        "train_score = np.zeros((n_iter,), dtype=np.float64)\n",
        "test_score = np.zeros((n_iter,), dtype=np.float64)\n",
        "for i, y_pred in enumerate(model.staged_predict(X_train)):\n",
        "    train_score[i] = mean_absolute_error(y_train, y_pred)\n",
        "for i, y_pred in enumerate(model.staged_predict(X_test)):\n",
        "    test_score[i] = mean_absolute_error(y_test, y_pred)\n",
        "\n",
        "fig = plt.figure(figsize=(6, 6))\n",
        "plt.subplot(1, 1, 1)\n",
        "plt.title('Deviance')\n",
        "plt.plot(np.arange(n_iter) + 1, train_score, 'b-',\n",
        "         label='Training Set Deviance')\n",
        "plt.plot(np.arange(n_iter) + 1, test_score, 'r-',\n",
        "         label='Test Set Deviance')\n",
        "plt.legend(loc='upper right')\n",
        "plt.xlabel('Boosting Iterations')\n",
        "plt.ylabel('Deviance')\n",
        "plt.ylim(0, 0.5)\n",
        "fig.tight_layout()\n",
        "plt.show()\n",
        "\n"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 441
        },
        "id": "4ivfdE2Wi3Bk",
        "outputId": "78667714-6895-4095-e54c-a0947b3ed761"
      },
      "source": [
        "from sklearn.ensemble import GradientBoostingClassifier\n",
        "X_synth = np.random.rand(1000, 3) * 2 - 1\n",
        "y_synth = np.sign(np.sign(X_synth[:, 2]) + np.sign(X_synth[:, 0] * X_synth[:, 1]) + 0.01)\n",
        "# The last feature is easy to track its contribution, but\n",
        "# the xor nature of the first two features can't be captured \n",
        "# by a depth-1 tree.\n",
        "\n",
        "X_tr, X_te, y_tr, y_te = train_test_split(X_synth, y_synth)\n",
        "\n",
        "gbc = GradientBoostingClassifier(max_depth=1, n_estimators=3000)\n",
        "gbc.fit(X_tr, y_tr)\n",
        "\n",
        "model = gbc\n",
        "n_iter = model.n_estimators_\n",
        "\n",
        "from sklearn.metrics import log_loss\n",
        "\n",
        "train_score = np.zeros((n_iter,), dtype=np.float64)\n",
        "test_score = np.zeros((n_iter,), dtype=np.float64)\n",
        "for i, y_pred in enumerate(model.staged_predict_proba(X_tr)):\n",
        "    train_score[i] = log_loss(y_tr, y_pred)\n",
        "for i, y_pred in enumerate(model.staged_predict_proba(X_te)):\n",
        "    test_score[i] = log_loss(y_te, y_pred)\n",
        "\n",
        "fig = plt.figure(figsize=(6, 6))\n",
        "plt.subplot(1, 1, 1)\n",
        "plt.title('Loss')\n",
        "plt.plot(np.arange(n_iter) + 1, train_score, 'b-',\n",
        "         label='Training Set loss')\n",
        "plt.plot(np.arange(n_iter) + 1, test_score, 'r-',\n",
        "         label='Test Set loss')\n",
        "plt.legend(loc='upper right')\n",
        "plt.xlabel('Boosting Iterations')\n",
        "plt.ylabel('LogLoss')\n",
        "fig.tight_layout()\n",
        "plt.show()\n"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LHnVdy8epXEL"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}