Partial Classification in the Belief Function Framework
Résumé
Partial, or set-valued classification assigns instances to sets of classes, making it possibleto reduce the probability of misclassification while still providing useful information. Thispaper reviews approaches to partial classification based on the Dempster-Shafer theory ofbelief functions. To define the utility of set-valued predictions, we propose to extend theutility matrix using an Ordered Weighted Average operator, allowing us to model the decisionmaker’s attitude towards imprecision using a single parameter. Various decision criteria areanalyzed comprehensively. In particular, two main strategies are distinguished: partialclassification based on complete preorders among partial assignments, and partial preordersamong complete assignments. Experiments with UCI and simulated Gaussian data sets showthe superiority of partial classification in terms of average utility, as compared to single-classassignment and classification with rejection.
Origine : Fichiers produits par l'(les) auteur(s)