import numpy as np
import math

def euclidian(test_case, train_case):
    ans = 0
    size = len(test_case)
    for i in range(0, size - 1):
        ans += (test_case[i] - train_case[i])**2
    return ans

def sortFirst(entry):
    return entry[0]

def predict(distances):
    classes = []
    for i in range(0, k):
        classes.append(distances[i][1])
    pred = np.bincount(classes).argmax()
    return pred

def knn(test_case):
    distances = []
    for train_case in train_dataset:
        ans = euclidian(test_case, train_case)
        distances.append([ans, train_case[90]])
    distances.sort(key=sortFirst)
    pred = predict(distances)
    if pred == test_case[90]:
        return True
    else:
        return False

data = np.loadtxt('./movement_libras.data', delimiter=',')
np.random.shuffle(data)
test_size = math.ceil(len(data)/10)
test_dataset = data[:test_size]
train_dataset = data[test_size:]
k = 7

correct = 0
wrong = 0
for test_case in test_dataset:
    if knn(test_case):
        correct += 1
    else:
        wrong += 1

print('Utilizando ', test_size, ' dados como teste em uma base de ', len(data), ' dados.')
print('Acertos: ', correct, '\nErros: ', wrong)