#!/usr/bin/env python
# coding: utf-8

# In[1]:


import numpy as np
from sklearn.datasets import load_iris
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier as skRandomForestClassifier


# In[2]:


def accuracy_score(y_true, y_pred):
    return np.mean(y_true == y_pred)


# In[3]:


class RandomForestClassifier():
    def __init__(self, n_estimators=100, max_depth=None, max_features="auto",
                 oob_score=False, random_state=0):
        self.n_estimators = n_estimators
        self.max_depth = max_depth
        self.max_features = max_features
        self.oob_score = oob_score
        self.random_state = random_state

    def fit(self, X, y):
        self.n_features_ = X.shape[1]
        self.classes_, y_train = np.unique(y, return_inverse=True)
        self.n_classes_ = len(self.classes_)
        MAX_INT = np.iinfo(np.int32).max
        rng = np.random.RandomState(self.random_state)
        self.estimators_ = []
        for i in range(self.n_estimators):
            est = DecisionTreeClassifier(max_depth=self.max_depth,
                                         max_features=self.max_features,
                                         random_state=rng.randint(MAX_INT))
            sample_rng = np.random.RandomState(est.random_state)
            sample_indices = sample_rng.randint(0, X.shape[0], X.shape[0])
            sample_counts = np.bincount(sample_indices, minlength=X.shape[0])
            est.fit(X, y_train, sample_weight=sample_counts)
            self.estimators_.append(est)
        if self.oob_score:
            self._set_oob_score(X, y_train)
        return self

    def _set_oob_score(self, X, y):
        predictions = np.zeros((X.shape[0], self.n_classes_))
        for i in range(self.n_estimators):
            sample_rng = np.random.RandomState(self.estimators_[i].random_state)
            sample_indices = sample_rng.randint(0, X.shape[0], X.shape[0])
            mask = np.ones(X.shape[0], dtype=bool)
            mask[sample_indices] = False
            predictions[mask] += self.estimators_[i].predict_proba(X[mask])
        self.oob_decision_function_ = predictions / np.sum(predictions, axis=1)[:, np.newaxis]
        self.oob_score_ = accuracy_score(y, np.argmax(predictions, axis=1))            

    def predict_proba(self, X):
        proba = np.zeros((X.shape[0], self.n_classes_))
        for i in range(self.n_estimators):
            proba += self.estimators_[i].predict_proba(X)
        proba /= self.n_estimators
        return proba

    def predict(self, X):
        proba = self.predict_proba(X)
        return self.classes_[np.argmax(proba, axis=1)]

    @property
    def feature_importances_(self):
        all_importances = np.zeros(self.n_features_)
        for i in range(self.n_estimators):
            all_importances += self.estimators_[i].feature_importances_
        all_importances /= self.n_estimators   
        return all_importances / np.sum(all_importances)


# In[4]:


X, y = load_iris(return_X_y=True)
clf1 = RandomForestClassifier(random_state=0, oob_score=True).fit(X, y)
clf2 = skRandomForestClassifier(random_state=0, oob_score=True).fit(X, y)
pred1 = clf1.predict(X)
pred2 = clf2.predict(X)
assert np.array_equal(pred1, pred2)
prob1 = clf1.predict_proba(X)
prob2 = clf2.predict_proba(X)
assert np.allclose(prob1, prob2)
assert np.allclose(clf1.oob_decision_function_, clf2.oob_decision_function_)
assert np.allclose(clf1.oob_score_, clf2.oob_score_)
assert np.allclose(clf1.feature_importances_, clf2.feature_importances_)