Towards Interpretable General Type-2 Fuzzy Classifiers

This paper presents two versions of a general type-2 fuzzy classifier. The focus is on interpretability since the rules are meaningful and the rule base is comprised of few rules, which is a direct consequence of the hierarchical reclassification process being proposed. The approaches are evaluated on a land cover classification problem by using data from a remote sensing platform. The classifiers´ performance are compared with the reference ones´ (maximum likelihood classifier and ordinary fuzzy classifier). The results show that the general type-2 fuzzy modeling is able to produce accurate classifiers while maintaining the model interpretability.