def regression(x_train_all, x_cv_all, targets_train, targets_cv, classifier):
clfs = []
mae_train = 0
mae_cv = 0
num_train = 0
num_cv = 0
for i in range(len(targets_train[0])):
x_train, y_train, x_cv, y_cv = feature_extraction.get_x_y_by_target(
x_train_all, x_cv_all, targets_train, targets_cv, i)
clf = classifier()
clf.fit(x_train, y_train)
clfs.append(clf)
p = clf.predict(x_cv)
mae_cv += utilities.ae(y_cv, p)
num_cv += len(y_cv)
p = clf.predict(x_train)
mae_train += utilities.ae(y_train, p)
num_train += len(y_train)
print 'Round %s completed.' % i
mae_train /= float(num_train)
mae_cv /= float(num_cv)
print 'MAE in training set: %s' % mae_train
print 'MAE in cv set: %s' % mae_cv
return clfs
def linear_regression_2(data):
print 'Training with linear regression 2...'
clf_map = {}
positions = [1, 2, 3, 4, 5, 10, 17, 24, 48, 72]
mae = 0.0
num = 0
for target in range(0, 39):
for pos in positions:
t = len(data[0]) - 39 + target
key = (target, pos)
x, y = feature_extraction.create_x_y(data)
clf = linear_model.LinearRegression()
clf.fit(x, y)
clf_map[key] = clf
p = clf.predict(x)
mae += utilities.ae(y, p)
num += len(y)
print '(%s, %s) completed.' % (target, pos)
mae /= float(num)
print 'MAE = %s' % mae
return clf_map