Research On Fuzzy Clustering Validity

In nowadays, society has become the information society. The informationgenerally exist on the data form, and the data is the carrier of information. So thedata for the information society is very important. A large number of data store in thedatabase, we need to use some certain methods for the classification. One of theimportant branch is clustering analysis. The basic principle of clustering analysis isthe same as the principle of brief "birds of a feather flock together". It is a processwhich classify data sample into a certain category through a certain attribute of thedata sample. The data in the same category object are the similar with each others,and different with other categories. Traditional clustering analysis method to theclassify data with more stringent requirements, an object usually can only belong toone category. However, in real objective circumstances, some objective object is tobelong to more than one kind of the nature, therefore the classification can't use therigid standard, but "soft classification". Hence fuzzy theory is introduced to theclustering analysis, using fuzzy theory can theoretically support the "softclassification". Usually, clustering analysis is in the classification state with nosupervision. So, How to evaluate cluster analysis of the classification whether accordwith the objective facts? In general, we use effectiveness evaluation, namely theeffectiveness of fuzzy clustering.In the fuzzy clustering, due to the validity of research is inconvenient, thereforegenerally with the best category for decision problem to the a lternative treatment.This paper summarizing the fuzzy clustering validity research history and currentresearch status firstly. And then discuss deeply on the effectiveness of the fuzzyclustering analysis indicators, There are mainly three kinds of indicators: Accordingto the statistics analysis of variance is put forward based on the effectiveness of theF statistic index, reflected in the close degree and the dispersion degree; In order toensure a higher degree of classification, highlight the influence of smaller statistics,and puts forward the effectiveness of mixed statistic index; According to thepromotion of the multi-dimensional data situation, based on sample data set categoryinternal statistical information and external statistical information to put forward theeffectiveness of false statistic index. To demonstrate the feasibility of effectiveness indexes, this paper applied a group of geological data carried on the detailedreasoning. The first use of fuzzy clustering analysis method of classification, andthen were applied to the analysis of multivariate F statistics, mixed statistics andthe effectiveness of the false statistic index to operate, analyze the change amountof data respectively, to draw the effectiveness of fuzzy clustering results. After usingthe comparison of the three kinds of effectiveness index, discriminate out the resultswhich more reasonable and more close to actual.