Research On Fuzzy Clustering Based On Weightingwith Cluster Center Separation

The fuzzy clustering is one of the key applications of data mining technology.It is the method of using fuzzy mathematics to study the clustering problem.Because it more accurately describes the uncertain relations in real problem,fuzzy clustering analysis has become research hot spot in the clustering analysis.Main contents are:1.Firstly,this paper introduces the background,theory development and research status of fuzzy clustering,as well as the frequently used fuzzy clustering algorithms.2.The basic knowledge of fuzzy set is introduced and several kinds of common fuzzy clustering algorithms are presented.FCM,FCM_CCS and WFCM_CCS are introduced in detail.3.Outfielders(points whose general behavior or characteristics are different from the most points in the sample concentration)can reduce the accuracy of clustering.Considering the influence of outfielders in clustering,this paper put forward a new algorithm that of fuzzy clustering based on sample-feature weights with cluster center separation.The model is IWFCM_CCS for short.The purpose of this model is to improve the accuracy of clustering for joining the sample weighting.Theoretical analysis on the model is made and an optimal solution is obtained.4.IWFCM_CCS is easy to program.Its algorithm complexity is similar to the FCM.In the paper,the algorithm's simulation experiment was carried out.It turned out that its accuracy is higher than IWFCM_CCS,much more higher than FCM,P FCM,SFPFKM and FCM_CCS.