Study Of FCM Algorithm On Parameters And Its Applications

Fuzzy c -Mean (FCM) clustering algorithm is one of the widely appliedalgorithms in unsupervised model recognition fields. As well known, the optimalsolution of FCM algorithm is obtained by minimizing the objective function, in whichtwo important parameters, fuzzy weighting exponent m and the number of clusters c,is involved. However, a poor partitioning result will be obtained for the improperchoices values of both m and c. Therefore the problem of optimization of both mand c is deserved to systematic study. For this purpose, based on partition fuzzy degree and fuzzy decision, a method ofthe optimal choice for parameter m is presented in this paper. Then by defining ofradius of data set to update the membership degree, a cluster validity function based onthe partition fuzzy degree is also proposed. Experimental results illustrate theeffectiveness of the optimal choice of m and c. At last, fuzzy c -mean clustering method is used to analyze microarray geneexpression data in order to detect differentially expressed gene. Compared with the resultsof model-based cluster analysis of microarray genes expression Data, our resultsindicate that fuzzy clustering can be a useful tool to exploit differential gene expressionfor microarray data.