import numpy as np
import random
from sklearn.decomposition import PCA
from sklearn.datasets import load_digits
from sklearn.svm import SVC
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

digits = load_digits()
m = len(digits.data)
x = digits.data
y = digits.target

num_components = 8

pca = PCA(n_components = num_components)
pca.fit(x)
x_transformed = pca.transform(x)

m_train = int(m*3/4)
m_test = m - m_train
print("Number of digit images:", m)
print("Train part:", m_train, " Test part:", m_test)

C = 4.0
kernel = "rbf"
num_repeats = 10
print("Number of repetitions:", num_repeats)

mean_accuracy = 0.
for i in range(num_repeats):
    x_train, x_test, y_train, y_test = train_test_split(
        x_transformed, y, test_size=1/4, shuffle=True, stratify=y
    )

    # model = SVC(C=C, kernel=kernel, degree=degree, coef0=coef0)
    model = SVC(C=C, kernel=kernel)
    model.fit(x_train, y_train)

    y_predicted = model.predict(x_test)
    accuracy = accuracy_score(y_true=y_test, y_pred=y_predicted)
    mean_accuracy += accuracy

mean_accuracy /= num_repeats
print("Mean accuracy of classification:", mean_accuracy)