Size reduction in fuzzy rulebases

A reconstruction methodology is proposed for dealing with size reduction in fuzzy rulebases. Elimination of redundant fuzzy rules is always at the root of size reduction. As a matter of fact, it is thus important to be aware of the different kinds of redundancy that may appear in fuzzy rulebases. Two types of redundancy are distinguished, i.e. interpolation redundancy and overlap redundancy. On the contrary to usual methods that are only able to deal with overlap redundancy, the presented method is efficient for both redundancy types. Furthermore, the proposed reconstruction principle induces the obtention of readable linguistic rules. The method performance is illustrated by means of two examples