# # print(y_pred)
# print('Final prediction score with optimal(use superknn): [%.8f]' % accuracy_score(y_test, y_pred))
'''方法2：使用need_proba,这时只有参数alpha和epsilon，其中alpha一共有models*classes = 4*10=40个,
grid_points = 10个
所以先调用opmimal_parameters函数先找到一个大概的范围'''
model = Superknn(models=models,
metric=accuracy_score,
regression=False,
random_state=0,
folds=5,
needs_proba=True,
shuffle=True)
# model.opmimal_parameters(X=X_train, y=y_train, eps_size=0.05, grid_points=10)
epsilons = np.arange(0.12, 0.16, 0.005)
model.alpha = 25
for epsilon in epsilons:
print('epsilon:', epsilon)
model.epsilon = epsilon
start_time = time.time()
model.fit(X_train, y_train)
print('fit took %fs!' % (time.time() - start_time))
start_time = time.time()
y_pred = model.predict(X_test)
print('predict took %fs!' % (time.time() - start_time))
# print(y_test)
# print(y_pred)
print('Final prediction score with optimal(use superknn): [%.8f]' %
accuracy_score(y_test, y_pred))
