# load data
path_to_data = "data/"
training, training_info, test, test_info = load_data(path_to_data)
# create model
nb_recipients_to_predict = 10
model = FrequencyModel(nb_recipients_to_predict=nb_recipients_to_predict)
# validation
n_split = 5
cross_validation = KFoldCrossValidation(n_split)
scores = []
for fold_nb, (training_fold, training_info_fold, test_fold, test_info_fold,
y_test_fold) in enumerate(
cross_validation.split(training, training_info)):
print("\nFold #%d..." % fold_nb)
# fit model
model.fit(training_fold, training_info_fold)
# predict
predictions_per_sender = model.predict(test_fold, test_info_fold)
# compute MAP@n
score = MAPn(predictions_per_sender, y_test_fold, nb_recipients_to_predict)
scores.append(score)
print("\nMAP@n for this fold -> %0.5f" % score)
# average scores over folds
final_score = np.mean(scores)
