KNN vs Logistic Regression vs SVM

import numpy
from sklearn import svm, linear_model, neighbors, datasets
import time


def predict():
    digit = datasets.load_digits()
    x = digit.data/digit.data.max()
    y = digit.target

    n = len(x)
    numpy.random.seed(int(time.time()))
    order = numpy.random.permutation(n)
    x = x[order]
    y = y[order]

    x_train = x[:int(.9*n)]
    y_train = y[:int(.9*n)]
    x_test = x[int(.9*n):]
    y_test = y[int(.9*n):]

    m_dict = {}
    nt = len(x_test)

    kernels = ['linear', 'poly', 'rbf']
    for kernel in kernels:
        svc = svm.SVC(kernel=kernel)
        svc.fit(x_train, y_train)
        y_pred = svc.predict(x_test)
        m_dict[kernel] = len(y_pred[y_test == y_pred])

    logistic = linear_model.LogisticRegression(multi_class='multinomial', solver='lbfgs')
    logistic.fit(x_train, y_train)
    y_pred =