Dynamic Clustering Method

Clustering analysis is one of the basic ways to understand things. The essential characters of things can be clearly recognized by clustering analysis. Up to now, dynamic clustering method is the most prevalent one in all sorts of clustering analysis methods. So the paper researches on dynamic clustering method mostly from two aspects-------static samples and dynamic samples.The general dynamic clustering algorithms are used by static samples. The results of clustering not only rely on the original classification, but easily get into local optimum. And the nearest neighbor clustering algorithm can just make up these disadvantages of dynamic clustering algorithms. However, the clustering radius of the present nearest neighbor clustering algorithm is selected randomly or confirmed by experience, and there is no correspondent cluster validity function to estimate the rationality of clustering results. Based on these reasons above, a new nearest neighbor clustering algorithm is proposed in this paper, and correspondent experimental analyses are also given out.In addition, the data is dynamic in a good many existing areas, such as data mining, large database and internet information processing, etc. Although ART2 neural network can well realize the clustering of dynamic samples, the problems arising from ART2 neural network itself restricted its application. Whereas, an improved ART2 neural network clustering algorithm is proposed to realize the clustering of dynamic samples, and the simulation results are given out at the same time.