import numpy as np
import time as t
from sklearn import svm, ensemble, naive_bayes, neighbors
from _image_classifier import ImageClassifier

def accuracy(y_true, y_pred):
    """returns the accuracy"""
    y_pred = np.round(y_pred)
    return (y_true == y_pred).mean()

def precision(y_true, y_pred):
    """returns the precision"""
    y_pred = np.round(y_pred)
    num = np.sum(np.logical_and(y_true, y_pred))
    den = np.sum(y_pred)
    return np.divide(num, den)

def recall(y_true, y_pred):
    """returns the recall"""
    y_pred = np.round(y_pred)
    num = np.sum(np.logical_and(y_true, y_pred))
    den = np.sum(y_true)
    return np.divide(num, den)

def f_measure(y_true, y_pred):
    """returns the F1 measure"""
    p = precision(y_true, y_pred)
    r = recall(y_true, y_pred)
    return 2 * p * r / (p + r)

def metric_test(iclf, metric, test_X, test_Y):
    return metric(test_Y, iclf.predict(test_X))

## Open data
im_train = np.load('Waldo_train_data.npy')
im_test = np.load('Waldo_test_data.npy')

lbl_train = np.load('Waldo_train_lbl.npy')
lbl_test = np.load('Waldo_test_lbl.npy')

# lbl_train = to_categorical(lbl_train)       # One hot encoding the labels
# lbl_test = to_categorical(lbl_test)

my_metric_test = lambda iclf, f: metric_test(iclf, f, im_test, lbl_test)

# ## Define model
svm_iclf = ImageClassifier(svm.SVC)
tree_iclf = ImageClassifier(neighbors.KNeighborsClassifier)
naive_bayes_iclf = ImageClassifier(naive_bayes.GaussianNB)
forest_iclf = ImageClassifier(ensemble.RandomForestClassifier)

classifiers = [
    svm_iclf,
    tree_iclf,
    naive_bayes_iclf,
    forest_iclf,
]

classifier_names = [
    "svm",
    "tree",
    "naive_bayes",
    "forest",
]

# print("name,time,accuracy,precision,recall,f_measure")
print("name,time,accuracy")
for clf, name in zip(classifiers, classifier_names):
    start = t.time()                            # Records time before training
    clf.fit(im_train, lbl_train)
    end = t.time()                              # Records time after tranining
    y_pred = clf.predict(im_test)
    print(
        name,
        end-start,
        clf.score(im_test, lbl_test),
        # precision(lbl_test, y_pred),
        # recall(lbl_test, y_pred),
        # f_measure(lbl_test, y_pred),
        sep=","
        )
    # print("training time:\t", end-start)
    # print("Accuracy:\t", clf.score(im_test, lbl_test))
    # print("Precision:\t", precision(lbl_test, y_pred))
    # print("Recall:\t\t", recall(lbl_test, y_pred))
    # print("F1-measure:\t", f_measure(lbl_test, y_pred))