def cv_knn(N_split, X_train, y_train, X_test, y_test, K_values):
X = np.concatenate((X_train, X_test), axis=0)
y = np.concatenate((y_train, y_test), axis=0)
# Splits
X_splits = np.split(X, N_split)
y_splits = np.split(y, N_split)
# Recherche
scores = {}
for K in K_values:
accuracys = []
for i in range(N_split):
X_train = np.concatenate(np.delete(X_splits, i, 0))
X_test = X_splits[i]
y_train = np.concatenate(np.delete(y_splits, i, 0))
y_test = y_splits[i]
model = Knn(K=K)
model.train(X_train, y_train)
evaluate = model.evaluate(X_test, y_test)
accuracys.append(evaluate['mean_accuracy'])
scores[K] = np.mean(accuracys)
# print + selection
print(max(scores.items(), key=operator.itemgetter(1)))
return max(scores.items(), key=operator.itemgetter(1))[0]
clf_Knn_abalone.train(X_train_abalone, y_train_abalone)
"""
Après avoir fait l'entrainement, évaluez votre modèle sur
les données d'entrainement.
IMPORTANT :
Vous devez afficher ici avec la commande print() de python,
- la matrice de confusion (confusion matrix)
- l'accuracy
- la précision (precision)
- le rappel (recall)
- le F1-score
"""
# Tester votre classifieur
print('---------- IRIS TRAIN ----------')
evaluate_train = clf_Knn_iris.evaluate(X_train_iris, y_train_iris)
for e in evaluate_train:
print(f'{e}\n {evaluate_train[e]}')
print('---------- WINE TRAIN ----------')
evaluate_train = clf_Knn_wine.evaluate(X_train_wine, y_train_wine)
for e in evaluate_train:
print(f'{e}\n {evaluate_train[e]}')
print('---------- ABALONE TRAIN ----------')
evaluate_train = clf_Knn_abalone.evaluate(X_train_abalone, y_train_abalone)
for e in evaluate_train:
print(f'{e}\n {evaluate_train[e]}')
"""
Finalement, évaluez votre modèle sur les données de test.
IMPORTANT :
