def wine_quality_prediction(selection_method, num_features):
X_train, X_test, y_train, y_test = prepare_wine_quality_data()
if selection_method == 'forward':
feature_set = forward_feature_selection(X_train, X_test, y_train, y_test, num_features)
elif selection_method == 'backward':
feature_set = backward_feature_selection_p_value(X_train, X_test, y_train, y_test, num_features)
# print(feature_set)
model = LinearRegression()
start = time.time()
model.fit(X_train[:, feature_set], y_train)
end = time.time()
time_diff = end - start
print(time_diff * 1000)
y_hat = model.predict(X_test[:, feature_set])
mean_square_error = mean_squared_error(y_test, y_hat)
return mean_square_error, time_diff*1000
def prediction(selection_method, num_features):
X_train, X_test, y_train, y_test = prepare_million_song_data()
if selection_method == 'forward':
feature_set = forward_feature_selection(X_train, X_test, y_train,
y_test, num_features)
elif selection_method == 'backward':
feature_set = backward_feature_selection_p_value(
X_train, X_test, y_train, y_test, num_features)
model = LinearRegression()
start = time.time()
model.fit(X_train[:, feature_set], y_train)
end = time.time()
time_diff = end - start
print(time_diff * 1000)
y_test_pred = model.predict(X_test[:, feature_set])
mean_error = calc_accuracy(y_test, y_test_pred)
print(mean_error)
# print(feature_set)
return mean_error, time_diff * 1000
def prediction(selection_method):
X_train, X_test, y_train, y_test = prepare_wine_quality_data()
if selection_method == 'forward':
feature_set = forward_feature_selection(X_train, X_test, y_train, y_test, 8)
elif selection_method == 'backward':
feature_set = backward_feature_selection_p_value(X_train, X_test, y_train, y_test, 8)
X_test, ids = prepare_wine_quality_test_data();
model = LinearRegression()
model.fit(X_train[:, feature_set], y_train)
y_test = model.predict(X_test[:, feature_set])
# print(y_test)
# print(feature_set)
return ids, y_test
def optimal_features(X_train, X_test, y_train, y_test,
feature_selection_method):
num_features = len(X_train[0])
model = LinearRegression()
num_feature_list = []
error_list = []
bic_list = []
for num in range(2, num_features + 1):
if feature_selection_method == 'forward':
feature_set = forward_feature_selection(X_train, X_test, y_train,
y_test, num)
elif feature_selection_method == 'backward':
feature_set = backward_feature_selection(X_train, X_test, y_train,
y_test, num)
# fit the model on training given the feature set
model.fit(X_train[:, feature_set], y_train)
mean_square_error = mean_squared_error(
y_test, model.predict(X_test[:, feature_set]))
sse = mean_square_error * len(y_test)
bic = num * math.log(len(y_test)) - 2 * math.log(sse)
print(feature_set, ' :: ', mean_square_error)
print(feature_set, ' :: ', bic)
num_feature_list.append(len(feature_set))
error_list.append(mean_square_error)
bic_list.append(bic)
return num_feature_list, error_list, bic_list