import numpy as np
import pandas
from sklearn.base import BaseEstimator, ClassifierMixin
from sklearn.datasets import make_classification
from sklearn.metrics import euclidean_distances, accuracy_score
from sklearn.model_selection import train_test_split
from sklearn.utils.multiclass import check_classification_targets
from sklearn.utils.validation import check_X_y, check_is_fitted, check_array
from math import sqrt
import math


class MyCustomClassifier(BaseEstimator, ClassifierMixin):

    def fit(self, X, y):
        X, y = check_X_y(X, y)
        check_classification_targets(y)
        X2 = []
        global mean2
        mean2 = []
        global std2
        std2 = []

        # set number of classes
        global classes
        classes = np.unique(y)
        print(classes)
        global features
        features = X.shape [1]
        print (X.shape)
        print(features)
        #X_1 = [[0 for col in range(features)] for row in range(len(classes))]

    # share between classes
        X2.append([])
        for i in range(len(classes)):
            X2.append([])
            for j in range(len(X)):
                X2.append([])
                if y[j] == i:
                    X2[i].append(X[j])

    # count mean and std

        #mean2.append([])
        #std2.append([])
        for i in range(len(classes)):
            mean2.append([])
            std2.append([])
            mean2[i].extend(np.mean(X2[i], axis=0))#extend
            std2[i].extend(np.std(X2[i], axis=0))


    def predict(self, X):
        normdistribution = []
        y_pred_X = []
        y_pred = []

# normal distribution
        #print (mean2[1][0])
        normdistribution.append([])
        for i in range(len(X)):
            normdistribution.append([])
            for j in range(len(classes)):
                normdistribution[i].append([])
                for k in range(features):
                    normdistribution[i][j].append([])
                    function = math.exp(-((X[i][k] - mean2[j][k]) ** 2) / (2 * (std2[j][k] ** 2)))/ (std2[j][k] * sqrt(2 * math.pi))
                    normdistribution[i][j][k] = function
        y_pred_X.append([])
        for i in range(len(X)):
            y_pred_X.append([])
            for j in range(len(classes)):
                y_pred_X[i].append([])
                for k in range(features):
                    if k == 0:
                        temp = 1
                    else:
                        temp = y_pred_X [i][j]
                    y_pred_X[i][j] = temp * normdistribution[i][j][k]

        global max

        for i in range(len(X)):
            max = 0
            for j in range(len(classes)-1):
                if y_pred_X[i][j] > y_pred_X[i][j+1]:
                    max = j
                else:
                    max = j+1

            y_pred.append(max)

        return y_pred
# Input validation
'''
X, y = make_classification(n_samples=1000, n_features=2, n_redundant=0, n_classes=2, random_state=0)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)

Gauss = MyCustomClassifier()
Gauss.fit(X_train, y_train)
y=Gauss.predict(X_test)

print("ACC %.3f in own implementation" % accuracy_score(y, y_test))

'''
from sklearn.utils.estimator_checks import check_estimator

check_estimator(MyCustomClassifier)