Split_vs_Bootstrap.ipynb

split_vs_bootstrap.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Split_vs_Bootstrap.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyMVtV3J6yM7jfaezPh3JqQX",
      "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/gist/alonsosilvaallende/cd9ec5ae0f6e8b83991606bd8e0f5848/split_vs_bootstrap.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cEwJ4mw3MRq8"
      },
      "source": [
        "!pip install -q scikit-survival"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "q8tsUA-yMY8i"
      },
      "source": [
        "from sksurv.datasets import load_gbsg2\n",
        "\n",
        "X, y = load_gbsg2()"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2vQfLWpMMcSp"
      },
      "source": [
        "scaling_cols = [c for c in X.columns if X[c].dtype.kind in ['i', 'f']]\n",
        "cat_cols = [c for c in X.columns if X[c].dtype.kind not in [\"i\", \"f\"]]"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yzZePe3PM1NT"
      },
      "source": [
        "from sklearn.compose import ColumnTransformer\n",
        "from sklearn.preprocessing import OrdinalEncoder\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "\n",
        "preprocessor = ColumnTransformer(\n",
        "    [('cat-preprocessor', OrdinalEncoder(), cat_cols),\n",
        "    ('standard-scaler', StandardScaler(), scaling_cols)],\n",
        "    remainder='passthrough', sparse_threshold=0)"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jdmci5AmTpFs"
      },
      "source": [
        "import numpy as np\n",
        "\n",
        "splits_seeds = np.random.RandomState(42).permutation(1000)[:10]"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BOQcJ9wTM9Bn"
      },
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "from sksurv.ensemble import RandomSurvivalForest\n",
        "from sksurv.metrics import concordance_index_censored\n",
        "\n",
        "ci_splits = []\n",
        "for seed in splits_seeds:\n",
        "  X_trn, X_test, y_trn, y_test = train_test_split(X, y, random_state=seed)\n",
        "  X_trn = preprocessor.fit_transform(X_trn)\n",
        "  X_test = preprocessor.transform(X_test)\n",
        "  rsf = RandomSurvivalForest()\n",
        "  rsf.fit(X_trn, y_trn)\n",
        "  ci_rsf = concordance_index_censored(y_test[\"cens\"], y_test[\"time\"], rsf.predict(X_test))\n",
        "  ci_splits.append(ci_rsf[0])"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qGGkvh54O8ip",
        "outputId": "8aba1b17-f1ff-45f4-cbc5-6626ecdc7f6d"
      },
      "source": [
        "np.mean(ci_splits)"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.6946017467642212"
            ]
          },
          "metadata": {},
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "czk3WPk6UwAK"
      },
      "source": [
        "# Bootstrap"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "E4BEIHWlQAPD"
      },
      "source": [
        "X_trn, X_test, y_trn, y_test = train_test_split(X, y, random_state=42)\n",
        "X_trn = preprocessor.fit_transform(X_trn)\n",
        "X_test = preprocessor.transform(X_test)"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TSMfhe--QDYy",
        "outputId": "4c56e32b-f3a4-4e97-d1f8-43d3b4479870"
      },
      "source": [
        "rsf = RandomSurvivalForest(random_state=42)\n",
        "rsf.fit(X_trn, y_trn)"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "RandomSurvivalForest(bootstrap=True, max_depth=None, max_features='auto',\n",
              "                     max_leaf_nodes=None, max_samples=None, min_samples_leaf=3,\n",
              "                     min_samples_split=6, min_weight_fraction_leaf=0.0,\n",
              "                     n_estimators=100, n_jobs=None, oob_score=False,\n",
              "                     random_state=42, verbose=0, warm_start=False)"
            ]
          },
          "metadata": {},
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XFnfFkrcQfSJ"
      },
      "source": [
        "bootstrap_seeds = np.random.RandomState(0).permutation(1000)[:10]"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YjHlvji6PNkg"
      },
      "source": [
        "bootstrap_indexes = \\\n",
        "[np.random.RandomState(_seed).choice(X_test.shape[0], X_test.shape[0], replace=True) for _seed in bootstrap_seeds]"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "s49YobPnShiZ"
      },
      "source": [
        "ci_bootstrap = \\\n",
        "[concordance_index_censored(y_test[bootstrap_indexes[0]][\"cens\"], \n",
        "                            y_test[bootstrap_indexes[0]]['time'], \n",
        "                            rsf.predict(X_test[bootstrap_indexes[0]]))[0] for i in range(len(bootstrap_indexes))]"
      ],
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "IfrKoPHfTWij",
        "outputId": "5df21d2a-3808-46ec-b874-a762b10e7221"
      },
      "source": [
        "np.mean(ci_bootstrap)"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.6869005705960908"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ]
    }
  ]
}