|
自动化学报 2005
一类基于数据的解释性模糊建模方法的研究, PP. 815-824 Keywords: Fuzzymodeling,interpretability,fuzzyclustering Abstract: ?Anapproachtoidentifyinterpretablefuzzymodelsfromdataisproposed.Interpretability,whichisoneofthemostimportantfeaturesoffuzzymodels,isanalyzedfirst.Thenumberoffuzzyrulesisdeterminedbyfuzzyclustervalidityindices.Amodifiedfuzzyclusteringalgorithm,combinedwiththeleastsquaremethod,isusedtoidentifytheinitialfuzzymodel.Anorthogonalleastsquarealgorithmandamethodofmergingsimilarfuzzysetsarethenusedtoremovetheredundancyofthefuzzymodelandimproveitsinterpretability.Next,inordertoattainhighaccuracy,whilepreservinginterpretability,aconstrainedLevenberg-Marquardtmethodisutilizedtooptimizetheprecisionofthefuzzymodel.Finally,theproposedapproachisappliedtoaPHneutralizationprocess,andtheresultsshowitsvalidity.
|