- All Implemented Interfaces:
- Serializable, ml.dmlc.xgboost4j.java.IEvaluation
public class MCCEvaluation2
extends Object
implements ml.dmlc.xgboost4j.java.IEvaluation
Evaluation function for Matthews Correlation Coefficient. For getting an
optimal MCC, a classification threshold needs to be used, which defines the
binary class assignment.
The most simplistic case would be to assume a threshold of 0.5.
However, this is in general not the best possible result. This evaluator
tries out different thresholds and selects the optimal threshold, which
yields the highest MCC. The highest MCC is then returned as evaluation result
- Author:
- pk
- See Also:
- Serialized Form