Selection of Clustering Algorithms Based on Grid-Connected Graph
挑选聚类算法的网格连通图方法

Each clustering algorithm has its suitable treatment of specific distribution data set. The SCGG method based on the Grid-Connected Graph is proposed to select suitable clustering algorithm for unknown distribution data set. The SCGG method analyses the potential structure of the data set. If the data set has ring clustering structure, the method select a single hierarchiral clustering algorithm, otherwise it selects the k-means. Experiment results show that SCGG is very efficient and successful.