snipet

from sklearn import metrics

np_test_x = np.arange(24)
np_test_y = dataset_T[5].values
print("hours\n", np_test_x)
print("traffic\n", np_test_y)

np_pred_y = np.polyval(poly_6, np_test_x)
print("predicted traffic\n", np_pred_y)
r_square = metrics.r2_score(np_test_y, np_pred_y)
print("R^2={}".format(r_square))


# 2차원 이상의 비선형 변수를 PolynomialFeatures 함수를 통해 선형으로 분해하여 LinearRegesssion 을 사용하는 방법

from sklearn.preprocessing import PolynomialFeatures
from sklearn.linear_model import LinearRegression
import numpy as np
import matplotlib.pyplot as plt

# 시간 변수: 0 ~ 23
np_data_hours = np.tile([int(hour) for hour in dataset.columns], dataset.values.shape[0]).reshape(-1, 1)
# 낮밤 변수: 0 (0~11), 1(12~23)
np_data_days = np.tile([int(int(hour) < 12) for hour in dataset.columns], dataset.values.shape[0]).reshape(-1, 1)

np_data_xs = np.concatenate((np_data_hours, np_data_days), axis = 1)
np_data_ys = dataset.values.reshape(-1, 1)
print(np_data_xs.shape)
print(np_data_ys.shape)

# train 과 test 나눔
idx = 120
np_train_xs = np_data_xs[:idx]
np_train_ys = np_data_ys[:idx]
np_test_xs = np_data_xs[idx:]
np_test_ys = np_data_ys[idx:]
print(np_train_xs.shape)
print(np_train_ys.shape)
print(np_test_xs.shape)
print(np_test_ys.shape)

np_test_xlabels = np_test_xs[:, 0]

# x feature transform
poly = PolynomialFeatures(degree=6)
np_train_xs = poly.fit_transform(np_train_xs)

# 학습
model = LinearRegression(fit_intercept=False)
model.fit(np_train_xs, np_train_ys)

# 예측
np_test_xs = poly.transform(np_test_xs)
np_pred_ys = model.predict(np_test_xs)

# 평가
plt.xticks(range(0, np_test_xlabels.shape[0]), np_test_xlabels)
plt.plot(np_pred_ys, c='r')
plt.plot(np_test_ys, c='b')
plt.show()

R_square = model.score(np_test_xs, np_test_ys)
print("R^2: {}".format(R_square))