Classification: objectives and metrics


Name	Used for optimization	User-defined parameters	Formula and/or description
Logloss	+	use_weights Default: true	Calculation principles $\text{[math]}$
CrossEntropy	+	use_weights Default: true	Calculation principles $\text{[math]}$
Precision	–	use_weights Default: true	Calculation principles $\text{[math]}$
Recall	–	use_weights Default: true	Calculation principles $\text{[math]}$
F1	–	use_weights Default: true	Calculation principles $\text{[math]}$
BalancedAccuracy	–	use_weights Default: true	Calculation principles $\text{[math]}$
BalancedErrorRate	–	use_weights Default: true	Calculation principles $\text{[math]}$
MCC	–	use_weights Default: true	Calculation principles $\text{[math]}$
Accuracy	–	use_weights Default: true	Calculation principles $\text{[math]}$
CtrFactor	–	use_weights Default: true	Calculation principles $\text{[math]}$
AUC *	–	use_weights Default: false	See the Wikipedia article.
BrierScore	–	–	Calculation principles $\text{[math]}$
HingeLoss	–	use_weights Default: true	Calculation principles $\text{[math]}$
HammingLoss	–	use_weights Default: true	Calculation principles $\text{[math]}$
ZeroOneLoss	–	use_weights Default: true	Calculation principles $\text{[math]}$
Kappa	–	–	Calculation principles $\text{[math]}$ $\text{[math]}$
WKappa	–	–	Calculation principles See the formula on page 3 of the A note on the linearly weighted kappa coefficient for ordinal scales paper.
LogLikelihoodOfPrediction	–	–	Calculation principles The calculation consists of the following steps: Define the sum of weights ( $\text{[math]}$ ) and the mean target ( $\text{[math]}$ ): $\text{[math]}$ $\text{[math]}$ Denote log-likelihood of a constant prediction: $\text{[math]}$ Calculate LogLikelihoodOfPrediction ( $\text{[math]}$ ), which reflects how the likelihood ( $\text{[math]}$ ) differs from the constant prediction: $\text{[math]}$


Name	Used for optimization	User-defined parameters	Formula and/or description
Logloss	+	use_weights Default: true	Calculation principles $\text{[math]}$
CrossEntropy	+	use_weights Default: true	Calculation principles $\text{[math]}$
Precision	–	use_weights Default: true	Calculation principles $\text{[math]}$
Recall	–	use_weights Default: true	Calculation principles $\text{[math]}$
F1	–	use_weights Default: true	Calculation principles $\text{[math]}$
BalancedAccuracy	–	use_weights Default: true	Calculation principles $\text{[math]}$
BalancedErrorRate	–	use_weights Default: true	Calculation principles $\text{[math]}$
MCC	–	use_weights Default: true	Calculation principles $\text{[math]}$
Accuracy	–	use_weights Default: true	Calculation principles $\text{[math]}$
CtrFactor	–	use_weights Default: true	Calculation principles $\text{[math]}$
AUC *	–	use_weights Default: false	See the Wikipedia article.
BrierScore	–	–	Calculation principles $\text{[math]}$
HingeLoss	–	use_weights Default: true	Calculation principles $\text{[math]}$
HammingLoss	–	use_weights Default: true	Calculation principles $\text{[math]}$
ZeroOneLoss	–	use_weights Default: true	Calculation principles $\text{[math]}$
Kappa	–	–	Calculation principles $\text{[math]}$ $\text{[math]}$
WKappa	–	–	Calculation principles See the formula on page 3 of the A note on the linearly weighted kappa coefficient for ordinal scales paper.
LogLikelihoodOfPrediction	–	–	Calculation principles The calculation consists of the following steps: Define the sum of weights ( $\text{[math]}$ ) and the mean target ( $\text{[math]}$ ): $\text{[math]}$ $\text{[math]}$ Denote log-likelihood of a constant prediction: $\text{[math]}$ Calculate LogLikelihoodOfPrediction ( $\text{[math]}$ ), which reflects how the likelihood ( $\text{[math]}$ ) differs from the constant prediction: $\text{[math]}$