Extraction of Fuzzy Rules from Data Including Qualitative Attributes Using Fuzzy Neural Networks with Forgetting
Kayo IMAMURA, K. Shinohara, Motohide Umano, Hiroyuki Tamura · Transactions of the Society of Instrument and Control Engineers · 1999
In this paper, we propose a method for extracting of fuzzy rules from data including qualitative attributes such as a country and an item type. These rules are extracted using fuzzy neural networks with forgetting, whose 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 functions using back propagation with forgetting and inverse of sigmoid function since the range of the membership functions must be in the unit interval [0, 1]. The proposed network is applied to sample data for estimation of human weight and real data for evaluation of system kitchen.