Discovering multiple fuzzy models using the EFMCN algorithm
Kalmanje S. Krishnakumar, A. Satyadas · 1995
Various qualitative and evolutionary fuzzy modeling techniques have been used by.resea.rchersfor IIOII linear system modeling.Evolutionary fuzzy modeling involves near opt,imal synt.hesis of fuzzy rules a.nd membership lunct.ionparameters using the genetic a.lgorit,hm optimization technique.However, in real life, several complex problems have mult.iple,equal or unequal, optimal solutions.This paper propose Evolutionary Fuzzy Modeling wit,h Clustered Niches (EFMCN) a.lgorithm to discover hierarchical clnst.erswhich may be used to ident.ifythese multiple fuzzy models.The niches are ident,ified using a sharing funct,ion which modifies t,he fit,ness of the individuals in a population based on a normalized distance measure, in conjunction with a centroid rlust ering algorit.hm.The main comp0nent.s of EFMCN are (;enrt.icAlgorit,hm (GA), Niching, Clustering, objective funct,ion.and Frizzy Syst,em (FS).The availabi1it.y of alter-na1.esolu!.ions provides the user wit.11 the rr~uch needed flexibility in ma.king cost effective choices.The paper concludes wit,h a. discussion on possible applications a.nd fut,ure directions.