The Research Of Fuzzy Clustering Based On Neural Network Model

We often need to solve the question of the Cluster with the development of the science and the advancement of the world. The cluster is an aged field. It ceaselessly progress along the development of the human society. Human being needs to distinguish one from aother and study the resemblance. We need the cluster in this work. Thus, the research of cluster analysis has a lot of important value in the theory, the project and the literal humaniores.The cluster algorithm has a lot of development, and gained plentiful and substantial result. Although many new algorithms appeared, we have not got a universal arithmetic. Some questions have not solved. The research of modern arithmetic algorithm will help to the development of cluster,The thesis bases on fuzzy theory. Systematically introduced the algorithm of fuzzy cluster face to the objective function. The research analyzes the similar standard, primitive data, cluster prototype that affects to cluster of the FCM calculate way creation, points out the advantage and the shortage of FCM calculate way in nowadays. This thesis put emphasis on discussing cluster network which based on fuzzy logic nerve, a simulated annealing algorithm for Cauchy training, the algorithm of FCM, it also improves and commixes arithmetic. It giving FCM arithmetic that bases on fuzzy logic nerve of cluster network and a simulated annealing with Cauchy training algorithm. Test verification that the validity of algorithm is very high. In computer calculation the validity of stochastic central test in cluster is up to 99%. It has the very nice performance.Classified the early work of the cluster analysis, commented on the sample data specification turns and the method of the distance computing, the degree of correlation proceeds. Introduced the fuzzy mathematics and the fuzzy cluster theory, analyzed the fuzzy theories that uses in the related research. Introduced to the fuzzy cluster method, analyzed the current international fuzzy cluster research progress.Based on fuzzy a variety of cluster "network algorithm proceeding, mainly researched in the algorithm to archetypal influence in cluster, cluster procedure energy, the cluster validity consumes with the cluster the etc. aspect does the analysis tallies up. At the same time, introduce the frequency parameter simulated annealing algorithm through the improvement, make the cluster network algorithm of the single and fuzzy nerve avoid sinking into the local smallest worth, then increases algorithm performance.Analyzed the system of the simulated annealing algorithm, introduce the Cauchy be trained ameliorative in order to improve simulated annealing algorithm. Change in the energy to the improvement algorithm, the cluster validity consumes with the cluster the time takes into the analysis. Putforward it with the nerve an algorithm hybrid, blending in order to improve the nerve an algorithm, reduce the sensitivity of the early cluster center. This thesis is divided into six chapters. In the first Chapter, discussing the basic theory of cluster and the foundation knowledge of the cluster; In the second chapter, introducing fuzzy mathematics, then introducing the misty cluster method, analyzing the current international research progress on fuzzy cluster; In the third chapter, discussing a misty cluster facing to the objective function, emphasize the FCM, In chapter 4, discussing the shortage of the nerve network algorithm in the fuzzy cluster application, give suggestion for the improvement; In chapter 5, proposing the algorithm of FCM which based on the cluster network of the fuzzy logic nerve and simulated annealing algorithm which with the training of Cauchy is validity that basal the algorithm of FCM, Test verification that the validity of algorithm is very high. The last chapter is a summary of the whole thesis. It also makes a prospect of our research.