Subsethood-based fuzzy modelling and classification

Reasoning with fuzzy rule-based models has been widely applied to perform various real world classification tasks. The main advantage of this approach is that it supports inferences in the way people think and make judgements. However, in order to gain high classification accuracy, transparency and interpretability of such models has often been ignored. To counter against this limitation, this paper proposes a quantifierbased fuzzy modelling method based on fuzzy subsethood measurements. The resulting induced models are transparent, interpretable whilst still able to provide high classification accuracy. This is confirmed by experimental comparative studies between this work and previous subsethood-based modelling approaches, for classification problems using benchmark datasets.