from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_absolute_error

test_y = np.loadtxt('testTarget.csv')
test_x  = np.loadtxt('testInput.csv')

def sci_RF(train_x, train_y, n_estimators, test_x, test_y):
	np.random.seed(400)
	np.random.shuffle(train_x)
	np.random.seed(400)
	np.random.shuffle(train_y)
	model = RandomForestRegressor(n_estimators, n_jobs=-1)
	model.fit(train_x, train_y)
	imp = model.feature_importances_
	print 'Train MAE: {}'.format(mean_absolute_error(train_y, model.predict(train_x)))
	print 'Test MAE: {}'.format(mean_absolute_error(test_y, model.predict(test_x)))
	return imp



imp = sci_RF(trainInput, trainTarget, 2500, test_x, test_y)
print(imp[idx])