stratified_group_k-fold.ipynb

{
 "metadata": {
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7-final"
  },
  "orig_nbformat": 2,
  "kernelspec": {
   "name": "python377jvsc74a57bd098b0a9b7b4eaaa670588a142fd0a9b87eaafe866f1db4228be72b4211d12040f",
   "display_name": "Python 3.7.7 64-bit ('base': conda)"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2,
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import random\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from collections import Counter, defaultdict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def stratified_group_k_fold(X, y, groups, k, seed=None):\n",
    "    labels_num = np.max(y) + 1\n",
    "    y_counts_per_group = defaultdict(lambda: np.zeros(labels_num))\n",
    "    y_distr = Counter()\n",
    "    for label, g in zip(y, groups):\n",
    "        y_counts_per_group[g][label] += 1\n",
    "        y_distr[label] += 1\n",
    "    \n",
    "    y_counts_per_fold = defaultdict(lambda: np.zeros(labels_num))\n",
    "    groups_per_fold = defaultdict(set)\n",
    "\n",
    "    def eval_y_counts_per_fold(y_counts, fold):\n",
    "        y_counts_per_fold[fold] += y_counts\n",
    "        std_per_label = []\n",
    "        for label in range(labels_num):\n",
    "            label_std = np.std([y_counts_per_fold[i][label] / y_distr[label] for i in range(k)])\n",
    "            std_per_label.append(label_std)\n",
    "        y_counts_per_fold[fold] -= y_counts\n",
    "        return np.mean(std_per_label)\n",
    "\n",
    "    groups_and_y_counts = list(y_counts_per_group.items())\n",
    "    random.Random(seed).shuffle(groups_and_y_counts)\n",
    "\n",
    "    for g, y_counts in sorted(groups_and_y_counts, key=lambda x: -np.std(x[1])):\n",
    "        best_fold = None\n",
    "        min_eval = None\n",
    "        for i in range(k):\n",
    "            fold_eval = eval_y_counts_per_fold(y_counts, i)\n",
    "            if min_eval is None or fold_eval < min_eval:\n",
    "                min_eval = fold_eval\n",
    "                best_fold = i\n",
    "        y_counts_per_fold[best_fold] += y_counts\n",
    "        groups_per_fold[best_fold].add(g)\n",
    "    \n",
    "    all_groups = set(groups)\n",
    "    for i in range(k):\n",
    "        train_groups = all_groups - groups_per_fold[i]\n",
    "        test_groups = groups_per_fold[i]\n",
    "\n",
    "        train_indices = [i for i, g in enumerate(groups) if g in train_groups]\n",
    "        test_indices = [i for i, g in enumerate(groups) if g in test_groups]\n",
    "\n",
    "        yield train_indices, test_indices\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "['female' 'female' 'male' ... 'male' 'male' 'male']\n"
     ]
    }
   ],
   "source": [
    "train_x = pd.read_csv('/opt/ml/input/data/train/train.csv')\n",
    "train_y = [train_x.gender.values]\n",
    "print(train_y)"
   ]
  }
 ]
}