指标参数metric
评价指标
{l2 for regression},
{binary_logloss for binary classification},
{multi_logloss for classification}
- “None”(字符串),表示无度量,等效于na,null,custom
- l1(absolute loss),等效于mean_absolute_error,mae,regression_l1
- l2(square loss),等效于mean_squared_error,mse,regression_l2,regression
- rmse(平方根损失),等效于root_mean_squared_error,l2_root
- quantile,分位数回归
- mape(MAPE损失),等效于mean_absolute_percentage_error
- huber(huber损失)
- fair(fair损失)
- poisson,泊松回归的负对数似然
- gamma,Gamma回归的负对数似然
- gamma_deviance,Gamma回归残差
- tweedie,Tweedie回归的负对数似然
- ndcg,等效于lambdarank,rank_xendcg,xendcg,xe_ndcg,xe_ndcg_mart,xendcg_mart
- map,等效于mean_average_precision
- auc
- average_precision,平均精度得分
- binary_logloss,log loss,等效于binary
- binary_error,对于单样本,0用于正确分类,1用于错误分类
- auc_mu
- multi_logloss,多分类log loss,等效于multiclass,softmax,multiclassova,multiclass_ova,ova,ovr
- multi_error,多分类错误率
- cross_entropy,交叉熵(具有可选的线性权重),等效于xentropy
- cross_entropy_lambda,“intensity-weighted”交叉熵,等效于 xentlambda
- kullback_leibler,Kullback-Leibler散度,等效于kldiv
支持多个指标,使用“,”分隔。
‘metric’: {‘binary_logloss’, ‘auc’}