{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn.datasets import load_iris\n",
    "from sklearn.dummy import DummyClassifier as skDummyClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "class DummyClassifier():\n",
    "    def __init__(self, strategy=\"stratified\", random_state=0, constant=None):\n",
    "        self.strategy = strategy\n",
    "        self.constant = constant\n",
    "        self.random_state = random_state\n",
    "\n",
    "    def fit(self, X, y):\n",
    "        self.classes_, y_train = np.unique(y, return_inverse=True)\n",
    "        self.n_classes_ = self.classes_.shape[0]\n",
    "        self.class_prior_ = np.bincount(y_train) / y.shape[0]\n",
    "        return self\n",
    "\n",
    "    def predict(self, X):\n",
    "        if self.strategy == \"most_frequent\" or self.strategy == \"prior\":\n",
    "            y = np.full(X.shape[0], self.classes_[np.argmax(self.class_prior_)])\n",
    "        elif self.strategy == \"constant\":\n",
    "            y = np.full(X.shape[0], self.constant)\n",
    "        elif self.strategy == \"uniform\":\n",
    "            rng = np.random.RandomState(self.random_state)\n",
    "            y = self.classes_[rng.randint(self.n_classes_, size=X.shape[0])]\n",
    "        elif self.strategy == \"stratified\":\n",
    "            y = self.classes_[np.argmax(self.predict_proba(X), axis=1)]\n",
    "        return y\n",
    "\n",
    "    def predict_proba(self, X):\n",
    "        if self.strategy == \"most_frequent\":\n",
    "            p = np.zeros((X.shape[0], self.n_classes_))\n",
    "            p[:, np.argmax(self.class_prior_)] = 1\n",
    "        elif self.strategy == \"prior\":\n",
    "            p = np.tile(self.class_prior_, (X.shape[0], 1))\n",
    "        elif self.strategy == \"constant\":\n",
    "            p = np.zeros((X.shape[0], self.n_classes_))\n",
    "            p[:, self.classes_ == self.constant] = 1\n",
    "        elif self.strategy == \"uniform\":\n",
    "            p = np.full((X.shape[0], self.n_classes_), 1 / self.n_classes_)\n",
    "        elif self.strategy == \"stratified\":\n",
    "            rng = np.random.RandomState(self.random_state)\n",
    "            p = rng.multinomial(1, self.class_prior_, size=X.shape[0])\n",
    "        return p"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "X, y = load_iris(return_X_y=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf1 = DummyClassifier(strategy=\"most_frequent\").fit(X, y)\n",
    "clf2 = skDummyClassifier(strategy=\"most_frequent\").fit(X, y)\n",
    "assert clf1.n_classes_ == clf2.n_classes_\n",
    "assert np.array_equal(clf1.classes_, clf2.classes_)\n",
    "assert np.array_equal(clf1.class_prior_, clf2.class_prior_)\n",
    "pred1 = clf1.predict(X)\n",
    "pred2 = clf2.predict(X)\n",
    "assert np.array_equal(pred1, pred2)\n",
    "prob1 = clf1.predict_proba(X)\n",
    "prob2 = clf2.predict_proba(X)\n",
    "assert np.array_equal(prob1, prob2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf1 = DummyClassifier(strategy=\"prior\").fit(X, y)\n",
    "clf2 = skDummyClassifier(strategy=\"prior\").fit(X, y)\n",
    "assert clf1.n_classes_ == clf2.n_classes_\n",
    "assert np.array_equal(clf1.classes_, clf2.classes_)\n",
    "assert np.array_equal(clf1.class_prior_, clf2.class_prior_)\n",
    "pred1 = clf1.predict(X)\n",
    "pred2 = clf2.predict(X)\n",
    "assert np.array_equal(pred1, pred2)\n",
    "prob1 = clf1.predict_proba(X)\n",
    "prob2 = clf2.predict_proba(X)\n",
    "assert np.array_equal(prob1, prob2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf1 = DummyClassifier(strategy=\"constant\", constant=0).fit(X, y)\n",
    "clf2 = skDummyClassifier(strategy=\"constant\", constant=0).fit(X, y)\n",
    "assert clf1.n_classes_ == clf2.n_classes_\n",
    "assert np.array_equal(clf1.classes_, clf2.classes_)\n",
    "assert np.array_equal(clf1.class_prior_, clf2.class_prior_)\n",
    "pred1 = clf1.predict(X)\n",
    "pred2 = clf2.predict(X)\n",
    "assert np.array_equal(pred1, pred2)\n",
    "prob1 = clf1.predict_proba(X)\n",
    "prob2 = clf2.predict_proba(X)\n",
    "assert np.array_equal(prob1, prob2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf1 = DummyClassifier(strategy=\"uniform\", random_state=0).fit(X, y)\n",
    "clf2 = skDummyClassifier(strategy=\"uniform\", random_state=0).fit(X, y)\n",
    "assert clf1.n_classes_ == clf2.n_classes_\n",
    "assert np.array_equal(clf1.classes_, clf2.classes_)\n",
    "assert np.array_equal(clf1.class_prior_, clf2.class_prior_)\n",
    "pred1 = clf1.predict(X)\n",
    "pred2 = clf2.predict(X)\n",
    "assert np.array_equal(pred1, pred2)\n",
    "prob1 = clf1.predict_proba(X)\n",
    "prob2 = clf2.predict_proba(X)\n",
    "assert np.array_equal(prob1, prob2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf1 = DummyClassifier(strategy=\"stratified\", random_state=0).fit(X, y)\n",
    "clf2 = skDummyClassifier(strategy=\"stratified\", random_state=0).fit(X, y)\n",
    "assert clf1.n_classes_ == clf2.n_classes_\n",
    "assert np.array_equal(clf1.classes_, clf2.classes_)\n",
    "assert np.array_equal(clf1.class_prior_, clf2.class_prior_)\n",
    "pred1 = clf1.predict(X)\n",
    "pred2 = clf2.predict(X)\n",
    "#assert np.array_equal(pred1, pred2)\n",
    "prob1 = clf1.predict_proba(X)\n",
    "prob2 = clf2.predict_proba(X)\n",
    "assert np.array_equal(prob1, prob2)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "dev",
   "language": "python",
   "name": "dev"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}