Python 72_Scikit_Learn을 이용한 Boston House Data 회귀분석2

회귀분석 응용

RM ~ LSTAT 두 변수를 이용한 다중회귀분석

# Price ~ RM + LSTAT + RM**2 + RM * LSTAT + LSTAT**2

# Price = b0 + b1 * rm + b2 * lstat + b3 * rm**2 + b4 * rm * lstat + b5 * lstat **2

# 학습 세트에 다항식항(컬럼)을 추가

X_train_rm_lstat_poly = poly.fit_transform(X_train_rm_lstat)

# 테스트 세트에 다항식항(컬럼)을 추가

X_test_rm_lstat_poly = poly.fit_transform(X_test_rm_lstat)

print(X_test_rm_lstat_poly[:2])

lin_reg.fit(X_train_rm_lstat_poly, y_train)

print(f'intercept: {lin_reg.intercept_}, coef: {lin_reg.coef_}')

[[  8.259      3.54      68.211081  29.23686   12.5316  ]

 [  6.312     10.58      39.841344  66.78096  111.9364  ]]

intercept: 58.131040227957655, coef: [-1.76285033e+01  1.52009093e+00  2.09295492e+00 -3.53889752e-01 -3.14275848e-03]

y_pred_rm_lstat_poly = lin_reg.predict(X_test_rm_lstat_poly)

mse = mean_squared_error(y_test, y_pred_rm_lstat_poly)

rmse = np.sqrt(mse)

r2 = r2_score(y_test, y_pred_rm_lstat_poly)

print(f'Price ~ RM + LSTAT: RMSE = {rmse}, R**2 = {r2}')

print('y true:', y_test[:5])

print('y pred:', y_pred_rm_lstat_poly[:5])

Price ~ RM + LSTAT: RMSE = 5.715001544053191, R**2 = 0.6303959336867977

y true: [42.8 21.2 31.  14.1 25.3]

y pred: [50.29506251 22.34374022 28.85971088 16.00625841 25.07303372]

RM ~ LSTAT 두 변수를 이용한 다중회귀분석2

# Price ~ RM + LSTAT + LSTAT**2

# Price = b0 + b1 * rm + b2 * lstat + b3 * lstat**2

X_train_last = np.c_[X_train_rm, X_train_lstat_poly]

X_test_last = np.c_[X_test_rm, X_test_lstat_poly]

print('X_train_last:', X_train_last[:2], '\n X_test_last: ', X_test_last[:2])

X_train_last: [[  5.093   29.68   880.9024]

                 [  6.251   16.44   270.2736]] 

 X_test_last:  [[  8.259    3.54    12.5316]

                 [  6.312   10.58   111.9364]]

lin_reg.fit(X_train_last, y_train) # fit/train

print(f'Price ~ RM + LSTAT + LSTAT**2: intercept: {lin_reg.intercept_}, coef {lin_reg.coef_}')

Price ~ RM + LSTAT + LSTAT**2: intercept: 11.976646227033507, coef [ 4.14148052 -1.79652146  0.03381396]

y_pred_last = lin_reg.predict(X_test_last) # 예측/테스트

print('y true:', y_test[:5])

print('y predict:', y_pred_last[:5].round(2))

y true: [42.8 21.2 31.  14.1 25.3]

y predict: [40.25 22.9  26.39 16.14 28.23]

mse = mean_squared_error(y_test, y_pred_last)

rmse = np.sqrt(mse)

r2 = r2_score(y_test, y_pred_last)

print(f'Price ~ RM + LSTAT: RMSE = {rmse}, R**2 = {r2}')

Price ~ RM + LSTAT: RMSE = 5.592903859241363, R**2 = 0.6460199841225782