The Research Of Data Mining Based On Fuzzy Cluster

Fuzzy clustering analysis is an important branch of unsupervised pattern recognition. Studying the algorithm and applications of fuzzy clustering is of great significance. Fuzzy clustering describes each sample's uncertainty and such uncertainty sometimes can reflect the real world better than other methods, so it becomes one of the mainstreams of clustering analysis.Because fuzzy clustering analysis is based on fuzzy set theory, this paper introduces the basic knowledge of fuzzy set theory, including the definition of fuzzy set, its basic theorem fuzzy relation and so on firstly. Then the research situation of fuzzy clustering analysis is studied, besides, this paper analyses the method of based on goal-function, discussing the evolvement of goal-function and the method of how to realize the algorithm. In all of the fuzzy clustering algorithms based on goal-function, Fuzzy c-means algorithm has perfect theory and wide application. So in chapter 3, the main thought of algorithm and the procedure of fuzzy c-means is studied in detail, meanwhile, the validity and existing problems now are also analyzed carefully. A new validity function is validated and the effect of weighted index m to this algorithm is also researched. At last, according to the existing problems of fuzzy c-means, we propose an improved FCM algorithm based on phased idea. In the first phase, the number of clustering and initialization of clustering center are dealt self-adaptively, and the better original parameters can be gained automatically by using fuzzy similar factor combined with F-statistic method. In the second phase, we adapt rival checked fuzzy c-means algorithm, which improves the method of computing the membership of the traditional FCM algorithm, to quicken constringency speed and enhance clustering effect. The improved FCM algorithm is validated by experiments and gets better result...