Weighting scheme for a pairwise multi-label classifier based on the fuzzy confusion matrix

In this work, we addressed the issue of improving the classification quality of label pairwise ensembles. Our goal is to improve the classification quality achieved by the ensemble via modification of the base classifiers that constitute the ensemble. To achieve this goal, a correction procedure that computes the measures of competence and cross-competence of a single classifier is proposed. These measures are used to modify the prediction of a base classifier. The measures are calculated using a dynamic confusion matrix. Additionally, we provide a weighting scheme that promotes the base classifiers that are the most susceptible to the correction based on the fuzzy confusion matrix. During the experimental study, the proposed approach was compared to two reference methods. The comparison was made in terms of eight different quality criteria. The result shows that the proposed method is able to improve classification quality when compared to baseline methods. c © 2018 Elsevier Ltd. All rights reserved.