Varför finjustera modellen?

Extreme Gradient Boosting med XGBoost

Sergey Fogelson

Head of Data Science, TelevisaUnivision

Exempel på ojusterad modell

import pandas as pd
import xgboost as xgb
import numpy as np
housing_data = pd.read_csv("ames_housing_trimmed_processed.csv")
X,y = housing_data[housing_data.columns.tolist()[:-1]],
        housing_data[housing_data.columns.tolist()[-1]]
housing_dmatrix = xgb.DMatrix(data=X,label=y)

untuned_params={"objective":"reg:squarederror"}

untuned_cv_results_rmse = xgb.cv(dtrain=housing_dmatrix,
        params=untuned_params,nfold=4,
        metrics="rmse",as_pandas=True,seed=123)

print("Untuned rmse: %f" %((untuned_cv_results_rmse["test-rmse-mean"]).tail(1)))

Untuned rmse: 34624.229980

Exempel på justerad modell

import pandas as pd
import xgboost as xgb
import numpy as np
housing_data = pd.read_csv("ames_housing_trimmed_processed.csv")
X,y = housing_data[housing_data.columns.tolist()[:-1]],
     housing_data[housing_data.columns.tolist()[-1]]
housing_dmatrix = xgb.DMatrix(data=X,label=y)

tuned_params = {"objective":"reg:squarederror",'colsample_bytree': 0.3,
    'learning_rate': 0.1, 'max_depth': 5}

tuned_cv_results_rmse = xgb.cv(dtrain=housing_dmatrix,
     params=tuned_params, nfold=4, num_boost_round=200, metrics="rmse",
     as_pandas=True, seed=123)

print("Tuned rmse: %f" %((tuned_cv_results_rmse["test-rmse-mean"]).tail(1)))

Tuned rmse: 29812.683594

Nu kör vi en övning!

Extreme Gradient Boosting med XGBoost