lab3 de la Bogdan [IA]

import numpy as np
from sklearn.naive_bayes import MultinomialNB
import matplotlib.pyplot as plt


# indexarea fiecarui interval
def value_to_bins(train_images, num_bins):
    return np.digitize(train_images, num_bins)



#calcularea scorului
def naive_bayes_score(training_data, training_labels, testing_data, testing_labels):
    naive_bayes_model = MultinomialNB()
    naive_bayes_model.fit(training_data, training_labels)
    naive_bayes_model.predict(testing_data)
    return naive_bayes_model.score(testing_data, testing_labels)

#returnarea y_pred
def naive_bayes_predict(training_data, training_labels, testing_data):
    naive_bayes_model = MultinomialNB()
    naive_bayes_model.fit(training_data, training_labels)
    return naive_bayes_model.predict(testing_data)


#afisarea imaginilor misclasate
def print_missclassified_images(y_pred, y_true, testing_data):
    i = 0
    for (x, y) in zip(y_pred, y_true):
        if(x != y):
            image = testing_data[i, :]
            image = np.reshape(image, (28, 28))
            plt.title("Aceasta imagine a fost clasificata ca " + str(x))
            plt.imshow(image.astype(np.uint8), cmap = 'gray')
            plt.show()
        i = i + 1


#afisarea matricei de confuzie
def confusion_matrix(y_true, y_pred):
    matrix = np.zeros((10, 10))
    for (x, y) in zip(y_pred, y_true):
        matrix[y.astype(int), x.astype(int)] = matrix[y.astype(int), x.astype(int)] + 1
    print(matrix)


train_images = np.loadtxt("data/train_images.txt")
test_images = np.loadtxt("data/test_images.txt")
train_labels = np.loadtxt("data/train_labels.txt")
test_labels = np.loadtxt("data/test_labels.txt")
num_bins = np.linspace(0, 255, 5)
#num_bins = [3, 5, 7, 9, 11]
digitized_train_images = value_to_bins(train_images, num_bins) - 1
digitized_test_images = value_to_bins(test_images, num_bins) - 1
print(naive_bayes_score(digitized_train_images, train_labels, digitized_test_images, test_labels))
y_pred = naive_bayes_predict(digitized_train_images, train_labels, digitized_test_images)
#print_missclassified_images(y_pred, test_labels, test_images)
confusion_matrix(test_labels, y_pred)