Acquisition of fuzzy rules from data including qualitative attributes using fuzzy neural networks with forgetting

We propose a method for extracting fuzzy rules from data including qualitative attributes such as countries and sex. These rules are extracted using fuzzy neural networks with forgetting, where membership functions for qualitative data are represented by enumerated fuzzy sets. We formulate them as switching units in fuzzy neural networks. We tune and prune these fuzzy neural networks using backpropagation with forgetting, where membership functions for qualitative attributes are updated by using the inverse of the sigmoid function since its ranges must be in the unit interval [0,1]. The proposed network is applied to sample data for estimating human weight and real data for evaluating system kitchens.