1、關於調參

調參是模型適應不同資料集的一個優化過程，如果只是建立了模型，而不對引數進行調整，是很不合理的。

2、xgboost調參

3、網路調參

from sklearn.metrics import fbeta_score, make_scorer,r2_score
from sklearn.model_selection import GridSearchCV

cv = KFold(n_splits=5,shuffle=True,random_state=45)  

parameters = {'alpha': [0.5,0.6,0.7]}  

clf=KernelRidge()

r2
 
 = make_scorer(r2_score)  
grid_obj = GridSearchCV(clf, parameters, cv=cv,scoring=r2)  
# grid_fit = grid_obj.fit(train, labels) 
grid_fit = grid_obj.fit(train_df.values, y_train_df)

best_clf = grid_fit.best_estimator_   

best_clf.fit(train_df.values, y_train_df) 

參考：