A Fu y Approach forKey ariables Identification
Yanfeng Hou, Waldemar Karwowski, William S. Marras · 2005
Identification ofinfluence ofinput variables isvery important forcomplex nonlinear systems withhighdimensional input space. Inthispaperwe propose a methodusing fuzzy average withfuzzy cluster distribution (FAFCD). Toavoid the interference ofdifferent distributions ofthesampling data, we dealwiththedistribution offuzzy clusters inthesampling data, instead oftheoriginal dataset.Todiscover theinput-output relationship, wefirst usemethodoffuzzy rules andFuzzyC- meanstopartition theoriginal sampling datasetintofuzzy clusters. Weproduce anewdatasetwiththesamedistribution of thefuzzy clusters. Thenthefuzzy average methodisapplied to thenewdataset. Bydoing this, theinterference ofdistribution of theoriginal sampling dataisremoved. Thismethodisstraight- forward andcomputationally easy. Theperformance istested on bothbenchmark dataandtheelectromyographic (EMG)signal Evaluation System.