Dynamic Classifier Aggregation Using Fuzzy t-conorm Integral

Fuzzy integral is a general aggregation operator,which encompasses many common aggregation operators likeweighted mean, ordered weighted mean, weighted minimumand maximum, etc. In classifier combining, it can be usedto aggregate the outputs of the individual classifiers in theteam with respect to a fuzzy measure, based on the classifierconfidences. In practice, the Choquet integral and the Sugenointegral are used most often. However, they both belong tothe more general family of fuzzy t-conorm integral. In thispaper, we theoretically examine which fuzzy t-conorm integralsare useful for classifier aggregation, and we experimentallycompare the individual methods on 23 benchmark datasets.