
!pip install kaggle 
from google.colab import files 
files.upload()

Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Requirement already satisfied: kaggle in /usr/local/lib/python3.7/dist-packages (1.5.12)
Requirement already satisfied: certifi in /usr/local/lib/python3.7/dist-packages (from kaggle) (2022.9.24)
Requirement already satisfied: six>=1.10 in /usr/local/lib/python3.7/dist-packages (from kaggle) (1.15.0)
Requirement already satisfied: python-slugify in /usr/local/lib/python3.7/dist-packages (from kaggle) (6.1.2)
Requirement already satisfied: urllib3 in /usr/local/lib/python3.7/dist-packages (from kaggle) (1.24.3)
Requirement already satisfied: python-dateutil in /usr/local/lib/python3.7/dist-packages (from kaggle) (2.8.2)
Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from kaggle) (2.23.0)
Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from kaggle) (4.64.1)
Requirement already satisfied: text-unidecode>=1.3 in /usr/local/lib/python3.7/dist-packages (from python-slugify->kaggle) (1.3)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->kaggle) (3.0.4)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->kaggle) (2.10)

Saving kaggle.json to kaggle.json

{'kaggle.json': b'{"username":"minaahayley","key":"3af569f6a8c028e6233a6d29ce2439d3"}'}


ls -1ha kaggle.json

kaggle.json


!mkdir -p ~/.kaggle #create folder name Kaggle
!cp kaggle.json ~/.kaggle #copy kaggle.jason into folder Kaggle
!chmod 600 ~/.kaggle/kaggle.json #ignore Permission Warning 

#다운로드가 제대로 되었는지 확인한다. 
#ls 명령어는 특정 경로에 어떤 파일이 있는지 확인해 보는 명령어다. 
%ls ~/.kaggle

kaggle.json


#Copy API command 후 데이터셋 다운로드하기  
!kaggle competitions download -c house-prices-advanced-regression-techniques

#파일 압축 풀기
!unzip house-prices-advanced-regression-techniques.zip

Downloading house-prices-advanced-regression-techniques.zip to /content
  0% 0.00/199k [00:00<?, ?B/s]
100% 199k/199k [00:00<00:00, 50.0MB/s]
Archive:  house-prices-advanced-regression-techniques.zip
  inflating: data_description.txt    
  inflating: sample_submission.csv   
  inflating: test.csv                
  inflating: train.csv


!mkdir -p ~/.kaggle/competitions/house-prices-advanced-regression-techniques #create folder name Kaggle
!cp sample_submission.csv ~/.kaggle/competitions/house-prices-advanced-regression-techniques #copy into folder Kaggle
!cp test.csv ~/.kaggle/competitions/house-prices-advanced-regression-techniques
!cp train.csv ~/.kaggle/competitions/house-prices-advanced-regression-techniques

#다운로드가 제대로 되었는지 확인한다. 
#ls 명령어는 특정 경로에 어떤 파일이 있는지 확인해 보는 명령어다. 
%ls ~/.kaggle/competitions/house-prices-advanced-regression-techniques

sample_submission.csv  test.csv  train.csv


import numpy as np 
import pandas as pd 
import seaborn as sns 
import matplotlib.pyplot as plt 

import warnings 
warnings.filterwarnings('ignore', category=RuntimeWarning)


submission = pd.read_csv('~/.kaggle/competitions/house-prices-advanced-regression-techniques/sample_submission.csv')
test = pd.read_csv('~/.kaggle/competitions/house-prices-advanced-regression-techniques/test.csv')
train = pd.read_csv('~/.kaggle/competitions/house-prices-advanced-regression-techniques/train.csv')


submission.sample()


# train 원본 데이터 저장 
train_org = train.copy()


# Null이 너무 많은 컬럼과 불필요한 컬럼 삭제 
train.drop(['Id', 'Alley', 'FireplaceQu', 'PoolQC', 'Fence', 'MiscFeature'], axis=1, inplace=True)


# 숫자형 Null 컬럼은 평균값으로 대체 
train.fillna(train.mean(), inplace=True)

/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:2: FutureWarning: Dropping of nuisance columns in DataFrame reductions (with 'numeric_only=None') is deprecated; in a future version this will raise TypeError.  Select only valid columns before calling the reduction.


# Null 값이 있는 피처명과 타입 추출 
train.isnull().sum()[train.isnull().sum()>0].index

Index(['MasVnrType', 'BsmtQual', 'BsmtCond', 'BsmtExposure', 'BsmtFinType1',
       'BsmtFinType2', 'Electrical', 'GarageType', 'GarageFinish',
       'GarageQual', 'GarageCond'],
      dtype='object')


# y 값의 분포 
train['SalePrice'].hist()

<matplotlib.axes._subplots.AxesSubplot at 0x7fde5c1eb310>


# 로그 변환 
train['SalePrice'] = np.log1p(train['SalePrice'])
train['SalePrice'].hist()

<matplotlib.axes._subplots.AxesSubplot at 0x7fde5c214b90>


from scipy.stats import skew 

# 숫자형 피처의 컬럼 index 추출 
feature_index = train.dtypes[train.dtypes != 'object'].index 
 
skew_features = train[feature_index].apply(lambda x : skew(x))

# skew(왜곡) 정도가 1이상인 컬럼만 추출 
skew_features_top = skew_features[skew_features > 1].sort_values(ascending=False)
skew_features_top

MiscVal          24.451640
PoolArea         14.813135
LotArea          12.195142
3SsnPorch        10.293752
LowQualFinSF      9.002080
KitchenAbvGr      4.483784
BsmtFinSF2        4.250888
ScreenPorch       4.117977
BsmtHalfBath      4.099186
EnclosedPorch     3.086696
MasVnrArea        2.673661
LotFrontage       2.382499
OpenPorchSF       2.361912
BsmtFinSF1        1.683771
WoodDeckSF        1.539792
TotalBsmtSF       1.522688
MSSubClass        1.406210
1stFlrSF          1.375342
GrLivArea         1.365156
dtype: float64


# 왜곡 정도가 높은 피처를 로그 변환
train[skew_features_top.index] = np.log1p(train[skew_features_top.index])


# 문자형 피처는 원 핫 인코딩으로 변환 
train = pd.get_dummies(train)
train.shape

(1460, 271)


# 원 핫 인코딩 후 Null 값 없음 
train.isnull().sum()[train.isnull().sum()>0].index

Index([], dtype='object')


# 회귀 계수 시각화 
def visualize_coefficient(models): 
    fig, axs = plt.subplots(figsize=(24,10), nrows=1, ncols=3)
    fig.tight_layout()

    for i, model in enumerate(models): 
        # 상위 10개, 하위 10개 회귀 계수를 구하고, 이를 판다스 concat으로 결합 
        coef = pd.Series(model.coef_, index=X.columns)
        coef_high = coef.sort_values(ascending=False).head(10)
        coef_low = coef.sort_values(ascending=False).tail(10)
        coef_concat = pd.concat([coef_high, coef_low])

        sns.barplot(x=coef_concat.values, y=coef_concat.index, ax=axs[i])


from sklearn.linear_model import LinearRegression, Ridge, Lasso 
from sklearn.model_selection import train_test_split 
from sklearn.metrics import mean_squared_error 

X = train.drop('SalePrice', axis=1) 
y = np.log1p(train['SalePrice'])
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=156)

# LinearRegression, Ridge, Lasso 학습, 예측, 평가 
lr_reg = LinearRegression()
ridge_reg = Ridge()
lasso_reg = Lasso()

lr_reg.fit(X_train, y_train)
ridge_reg.fit(X_train, y_train)
lasso_reg.fit(X_train, y_train)

models = [lr_reg, ridge_reg, lasso_reg]
visualize_coefficient(models)


plt.scatter(x=train_org['GrLivArea'], y=train_org['SalePrice'])

<matplotlib.collections.PathCollection at 0x7fde571b4250>


# GrLivArea, SalePrice 모두 로그 변환됐으므로 이를 반영한 조건 생성 
outlier_index = train[(train['GrLivArea'] > np.log1p(4000)) & (train['SalePrice'] > np.log1p(50000))].index

# 이상치 데이터 삭제 
train.drop(outlier_index, axis=0, inplace=True)


# 로그 변환 된 rmse를 측정하는 함수 생성 
def get_rmse(model):
    pred = model.predict(X_test)
    mse = mean_squared_error(y_test, pred)
    rmse = np.sqrt(mse)
    print(model.__class__.__name__, '로그 변환된 rmse:', np.round(rmse,3))
    return rmse


# 여러 모델의 rmse 값을 반환 
def get_rmses(models):
    rmses = []
    for model in models:
        rmse = get_rmse(model)
        rmses.append(rmse)
    return rmses


from sklearn.linear_model import LinearRegression, Ridge, Lasso 
from sklearn.model_selection import train_test_split 
from sklearn.metrics import mean_squared_error 

X = train.drop('SalePrice', axis=1) 
y = train['SalePrice'] 
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=156)

# LinearRegression, Ridge, Lasso 학습, 예측, 평가 
lr_reg = LinearRegression()
ridge_reg = Ridge()
lasso_reg = Lasso()

lr_reg.fit(X_train, y_train)
ridge_reg.fit(X_train, y_train)
lasso_reg.fit(X_train, y_train)

models = [lr_reg, ridge_reg, lasso_reg]
get_rmses(models)

LinearRegression 로그 변환된 rmse: 0.132
Ridge 로그 변환된 rmse: 0.125
Lasso 로그 변환된 rmse: 0.258

[0.13195207393230912, 0.12459727425511735, 0.2575777799483299]


from sklearn.model_selection import GridSearchCV 

def best_params(model, params): 
    grid_model = GridSearchCV(model, param_grid=params, scoring='neg_mean_squared_error', cv=5)
    grid_model.fit(X, y)
    rmse = np.sqrt(-1 * grid_model.best_score_) 
    print('{0} 5 CV 시 최적 평균 rmse: {1}, 최적 평균 alpha: {2}'.format(model.__class__.__name__, np.round(rmse, 4), grid_model.best_params_))

ridge_params = {'alpha' : [0.05, 0.1, 1, 5, 8, 10, 12, 15, 20]}
lasso_params = {'alpha' : [0.001, 0.005, 0.008, 0.05, 0.03, 0.1, 0.5, 1, 5, 10]}

best_params(ridge_reg, ridge_params)
best_params(lasso_reg, lasso_params)

Ridge 5 CV 시 최적 평균 rmse: 0.1125, 최적 평균 alpha: {'alpha': 8}
Lasso 5 CV 시 최적 평균 rmse: 0.1121, 최적 평균 alpha: {'alpha': 0.001}


# 앞의 최적화 alpha 값으로 학습, 예측, 평가 
lr_reg = LinearRegression()
ridge_reg = Ridge(alpha=8)
lasso_reg = Lasso(alpha=0.001)

lr_reg.fit(X_train, y_train)
ridge_reg.fit(X_train, y_train)
lasso_reg.fit(X_train, y_train)

models = [lr_reg, ridge_reg, lasso_reg]
get_rmses(models)

LinearRegression 로그 변환된 rmse: 0.132
Ridge 로그 변환된 rmse: 0.117
Lasso 로그 변환된 rmse: 0.116

[0.13195207393230912, 0.11736730851679039, 0.1158658441187506]


# 피처 중요도 시각화 
models = [lr_reg, ridge_reg, lasso_reg]
visualize_coefficient(models)


from sklearn.ensemble import GradientBoostingRegressor 
from xgboost import XGBRegressor 
from lightgbm import LGBMRegressor 

gbm_reg = GradientBoostingRegressor(n_estimators=1000)
xgb_reg = XGBRegressor(n_estimators=1000, learning_rate=0.05, colsample_bytree=0.5, subsample=0.8)
lgbm_reg = LGBMRegressor(n_estimators=1000, learning_rate=0.05, num_leaves=4, 
                         subsample=0.6, colsample_bytree=0.4, reg_lambda=10, n_jobs=-1)

gbm_reg.fit(X_train, y_train)
xgb_reg.fit(X_train, y_train)
lgbm_reg.fit(X_train, y_train)

models = [gbm_reg, xgb_reg, lgbm_reg]
get_rmses(models)

[07:42:56] WARNING: /workspace/src/objective/regression_obj.cu:152: reg:linear is now deprecated in favor of reg:squarederror.
GradientBoostingRegressor 로그 변환된 rmse: 0.124
XGBRegressor 로그 변환된 rmse: 0.116
LGBMRegressor 로그 변환된 rmse: 0.118

[0.12434554514384757, 0.11569128150068062, 0.11806601740834839]


# 개별 모델의 학습 
ridge_reg = Ridge(alpha=8)
lasso_reg = Lasso(alpha=0.001)
ridge_reg.fit(X_train, y_train)
lasso_reg.fit(X_train, y_train)

# 개별 모델 예측 
ridge_pred = ridge_reg.predict(X_test)
lasso_pred = lasso_reg.predict(X_test)

# 개별 모델 예측값 혼합으로 최종 예측값 도출 
pred = 0.4 * ridge_pred + 0.6 * lasso_pred 

# 최종 혼합 모델의 rmse 값 출력 
mse = mean_squared_error(y_test, pred) 
rmse = np.sqrt(mse)
print('로그 변환된 rmse:', np.round(rmse,3))

로그 변환된 rmse: 0.115


# 개별 모델의 학습 
xgb_reg = XGBRegressor(n_estimators=1000, learning_rate=0.05, colsample_bytree=0.5, subsample=0.8)
lgbm_reg = LGBMRegressor(n_estimators=1000, learning_rate=0.05, num_leaves=4, 
                         subsample=0.6, colsample_bytree=0.4, reg_lambda=10, n_jobs=-1)
xgb_reg.fit(X_train, y_train)
lgbm_reg.fit(X_train, y_train)

# 개별 모델 예측 
xgb_pred = xgb_reg.predict(X_test)
lgbm_pred = lgbm_reg.predict(X_test)

# 개별 모델 예측값 혼합으로 최종 예측값 도출 
pred = 0.5 * xgb_pred + 0.5 * lgbm_pred 

# 최종 혼합 모델의 rmse 값 출력 
mse = mean_squared_error(y_test, pred) 
rmse = np.sqrt(mse)
print('로그 변환된 rmse:', np.round(rmse,3))

[07:44:06] WARNING: /workspace/src/objective/regression_obj.cu:152: reg:linear is now deprecated in favor of reg:squarederror.
로그 변환된 rmse: 0.114

티스토리

house prices advanced regression techniques

Kaggle과 Colab 연동¶

Feature Engineering¶

데이터 분포도¶

원 핫 인코딩¶

이상치 데이터 처리¶

선형 회귀¶

모델 학습/예측/평가¶

하이퍼 파라미터 튜닝¶

회귀 트리¶

혼합 모델¶