게으른 우루루

Ridge 참조

• Ridge

Random Forest 참조

• Random Forest -- intro

• Random Forest

### load library

from sklearn.datasets import load_boston

from sklearn.model_selection import train_test_split

from sklearn.preprocessing import MinMaxScaler, OneHotEncoder, PolynomialFeatures

from sklearn.linear_model import Ridge

from sklearn.ensemble import RandomForestRegressor

### data load

boston = load_boston() 

# print(boston.keys())

x_train, x_test, y_train, y_test = \

  train_test_split(boston.data, boston.target,

                   test_size=0.3, random_state=0)

### scale 조정

scaler_mms = MinMaxScaler().fit(x_train) # scaler 호출

x_train_scaled = scaler_mms.transform(x_train) # scaler를 x_train에 적용

x_test_scaled = scaler_mms.transform(x_test) # scaler를 x_test를 적용

### model

randfor = RandomForestRegressor(n_estimators=100, random_state=0, n_jobs=-1) # n_jobs = 코어의 수(-1 : all)

ridge = Ridge()

models = [randfor, ridge]

### model 적용 및 성능 테스트

for degree in range(1, 6):

    poly = PolynomialFeatures(degree=degree, include_bias=False).fit(x_train_scaled)# include_bias = True: x_train_scaled 맨 왼쪽에 상수항 추가. 

    x_train_poly = poly.transform(x_train_scaled)

    x_test_poly = poly.transform(x_test_scaled)

    print('degree={}'.format(degree))

    print('x_train.shape={}'.format(x_train.shape))

    print('x_train_poly.shape={}\n'.format(x_train_poly.shape))

    print('원래 특성과 polynomial 한 특성의 수 \n==> origin:{}, poly:{}\n'.format(

        boston.data.shape[1], len(poly.get_feature_names())))

    for model in models:

        conf = model.fit(x_train_scaled, y_train)

        print('교차항이 없을 때 {} 점수: {:.3f}'.format(

            model.__class__.__name__, model.score(x_test_scaled, y_test))) # __class__.__name__ : class이름 호출

        conf = model.fit(x_train_poly, y_train)

        print('degree={}, 교차항이 있을 때 {} 점수{:.3f}\n'.format(

            degree, model.__class__.__name__, model.score(x_test_poly, y_test)))

    ptrint('----------------------------------------------')

모델 성능 테스트 평가의 일부분