机器学习中评价指标的选择

一、分类

    二分类：eval_metric='auc'/'logloss'

    多分类：eval_metric='mlogloss'

    1、样本均衡：

        准确度，二分类还可以选择auc。

from sklearn.metrics import accuracy_score
from sklearn.metrics import classification_report
from sklearn.metrics import roc_auc_score
print(roc_auc_score(y_test,test_pre)) #test_pre为概率值
print(accuracy_score(y_test,test_pre2))
print(classification_report(y_test, test_pre2))

    2、样本不均衡

        f1值

from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
print(classification_report(y, predictions))
print(confusion_matrix(y, predictions))

二、回归

    eval_metric='rmse'

 

    r2、rmse、mae

from sklearn.metrics import r2_score
print(r2_score(y_train, train_pre))
print(r2_score(y_valid, valid_pre))

from sklearn.metrics import mean_squared_error
print(math.sqrt(mean_squared_error(y_train, train_pre)))
print(math.sqrt(mean_squared_error(y_valid, valid_pre)))

mean_absolute_error(np.exp(y), np.exp(yhat)