In this work we addressed the issue of applying a stochastic classifier and alocal, fuzzy confusion matrix under the framework of multi-labelclassification. We proposed a novel solution to the problem of correcting labelpairwise ensembles. The main step of the correction procedure is to computeclassifier-specific competence and cross-competence measures, which estimateserror pattern of the underlying classifier. At the fusion phase we employed twoweighting approaches based on information theory. The classifier weightspromote base classifiers which are the most susceptible to the correction basedon the fuzzy confusion matrix. During the experimental study, the proposedapproach was compared against two reference methods. The comparison was made interms of six different quality criteria. The conducted experiments reveals thatthe proposed approach eliminates one of main drawbacks of the originalFCM-based approach i.e. the original approach is vulnerable to the imbalancedclass/label distribution. What is more, the obtained results shows that theintroduced method achieves satisfying classification quality under allconsidered quality criteria. Additionally, the impact of fluctuations of dataset characteristics is reduced.
展开▼
