Adaptive Clustering Algorithm Based On The Density Parameters Of The Model

Self-adaptation is vital for various applications of clustering algorithms, while at the same time, the efforts that address this issue are far from enough. This dissertation aims at improving the self-adaptation of clustering algorithm without sacrificing the efficiency and effect of the clustering algorithm.In this dissertation, the clustering mechanism of the typical algorithms is profoundly studied, based on which, it is demonstrated that it is necessary to make the modeling of data summarizing, similarity measuring and clustering mechanism in an integrated way. By analyzing the merits and defects of the density based clustering algorithm, the concept of density model is proposed. After profound study of the informative nearest neighborhood and its indication of the boundary of clusters, a dynamically constructed nearest neighbor graph is proposed to find different location model of points. A parameter-free clustering algorithm based on density model is proposed in this dissertation. It requires neither previously nor interactively setting of pivotal parameters via range scaling and proportional criterion technique.The theoretical analysis shows that time complexity of the proposed algorithm is relatively low. And the experimental results demonstrate that the proposed algorithm can correctly recognize the arbitrary shaped clusters.