Supplemental_Analyses.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%config InlineBackend.figure_formats = ['svg']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import random\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from joblib import dump, load\n",
    "np.random.seed(42)\n",
    "random.seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir = './data/'\n",
    "with np.load('data/Xy.npz') as f:\n",
    "    X = f['X']\n",
    "    y = f['y']\n",
    "    y34 = f['y34']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Perform temporal split of data into train/test sets\n",
    "pop = pd.read_csv(data_dir + 'population/d10_with_vitals.csv').set_index('BMT_ID')\n",
    "\n",
    "split_date = 201701001\n",
    "split_idx = -85\n",
    "\n",
    "assert (pop[:split_idx].index < split_date).all()\n",
    "assert (pop[split_idx:].index >= split_date).all()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import preprocessing, model_selection, metrics, utils\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from tqdm import tqdm\n",
    "from joblib import Parallel, delayed\n",
    "from sklearn.base import clone"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Specify hyperparameters and cv parameters\n",
    "base_estimator = LogisticRegression(penalty='l2', class_weight='balanced', solver='liblinear')\n",
    "param_grid = {\n",
    "    'C': [10. ** n for n in range(-6, 7)],\n",
    "    'penalty': ['l2'],\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Alternative label definition\n",
    "{0,1,2} -> negative, {3,4} -> positive"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 472,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.31790123456790126, 0.13580246913580246)"
      ]
     },
     "execution_count": 472,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.mean(), y34.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 376,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/data4/tangsp/venv/lib/python3.7/site-packages/sklearn/model_selection/_search.py:814: DeprecationWarning: The default of the `iid` parameter will change from True to False in version 0.22 and will be removed in 0.24. This will change numeric results when test-set sizes are unequal.\n",
      "  DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "Xtr, Xte = X[:split_idx], X[split_idx:]\n",
    "ytr, yte = y34[:split_idx], y34[split_idx:]\n",
    "\n",
    "cv_splits, cv_repeat = 5, 20\n",
    "cv = model_selection.RepeatedStratifiedKFold(cv_splits, cv_repeat, random_state=0)\n",
    "clf = model_selection.GridSearchCV(\n",
    "    clone(base_estimator), param_grid, \n",
    "    cv=cv, scoring='roc_auc', n_jobs=5,\n",
    ")\n",
    "clf.fit(Xtr, ytr)\n",
    "test_score = metrics.roc_auc_score(yte, clf.decision_function(Xte))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 377,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "                                                  \r"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test AUC: 0.596 (0.423, 0.761)\n"
     ]
    }
   ],
   "source": [
    "y_true = yte\n",
    "y_score = clf.decision_function(Xte)\n",
    "\n",
    "def boostrap_func(i, y_true, y_score):\n",
    "    yte_true_b, yte_pred_b = utils.resample(y_true, y_score, replace=True, random_state=i)\n",
    "    return metrics.roc_curve(yte_true_b, yte_pred_b), metrics.roc_auc_score(yte_true_b, yte_pred_b)\n",
    "\n",
    "roc_curves, auc_scores = zip(*Parallel(n_jobs=4)(delayed(boostrap_func)(i, y_true, y_score) for i in tqdm(range(1000), leave=False)))\n",
    "print('Test AUC: {:.3f} ({:.3f}, {:.3f})'.format(np.median(auc_scores), np.percentile(auc_scores, 2.5), np.percentile(auc_scores, 97.5)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 378,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import RandomizedSearchCV\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "import scipy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 379,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=<sklearn.model_selection._split.RepeatedStratifiedKFold object at 0x7f9b8ffacc50>,\n",
       "                   error_score='raise-deprecating',\n",
       "                   estimator=RandomForestClassifier(bootstrap=True,\n",
       "                                                    class_weight=None,\n",
       "                                                    criterion='gini',\n",