인프런 커뮤니티 질문&답변

# 라이브러리 및 데이터 불러오기
import pandas as pd
train = pd.read_csv("https://raw.githubusercontent.com/lovedlim/inf/refs/heads/main/p4/8_2/churn_train.csv")
test = pd.read_csv("https://raw.githubusercontent.com/lovedlim/inf/refs/heads/main/p4/8_2/churn_test.csv")

#EDA
print(train.shape, test.shape)
#(4116, 19) (1764, 18)
# print(train.info())
# print(test.info())
#결측치 x

# print(train.isnull().sum())
# print(test.isnull().sum())

#,수치형,오브젝트 체크
# train.describe(include='O')
# train.describe()
#Id는 버리기
# train['TotalCharges'].value_counts() #타겟이 많으면 회귀

#데이터 전처리
train = train.drop('customerID',axis=1)
test=  test.drop('customerID',axis =1)
print(train.shape, test.shape)
#타겟 분리
target = train.pop('TotalCharges')

#스케일링
#컬럼다르면 합치고 인코딩후 분리
df = pd.concat([train,test],axis=0)
train1 = df.iloc[:len(train)].copy()
test1 = df.iloc[len(train):].copy()
print(train.shape, test.shape)
from sklearn.preprocessing import RobustScaler
cols = train1.describe(exclude='object').columns
scaler = RobustScaler()
train1[cols] = scaler.fit_transform(train1[cols])
test1[cols] = scaler.transform(test1[cols])

#인코딩
from sklearn.preprocessing import LabelEncoder
cols = train1.describe(include='object').columns
le = LabelEncoder()
for col in cols:
  train1[col] = le.fit_transform(train1[col])
  test1[col] = le.transform(test1[col])
print(train.shape, test.shape)

#검증데이터 분리
from sklearn.model_selection import train_test_split
x_tr,x_val,y_tr,y_val = train_test_split(train1, target, test_size=0.2, random_state=0)
print(x_tr.shape,x_val.shape,y_tr.shape,y_val.shape)

#모델 학습(랜덤)
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_absolute_error
rf = RandomForestRegressor(random_state=0)
rf.fit(x_tr,y_tr)
pred = rf.predict(x_val)
print(mean_absolute_error(y_val,pred))

#lightGBM
import lightgbm as lgb
lgbmr = lgb.LGBMRegressor(random_state=0, verbose=0)
lgbmr.fit(x_tr,y_tr)
pred = lgbmr.predict(x_val)
print(mean_absolute_error(y_val,pred))

#xgboost
from xgboost import XGBRegressor
xgbr =  XGBRegressor(random_state=0)
xgbr.fit(x_val,y_val)
pred = xgbr.predict(x_val)
print(mean_absolute_error(y_val,pred))
#951.033800718042 랜덤
#952.79254077987 lightgbm
#41.89787395866172 xgboost

맨마지막에 xgboost로 돌렸는데 41이라는 결과 나옵니다 이건 과적합인가요?