Minimum Spanning Tree Clustering Algorithm
Hou Dong-feng · Journal of Chinese Computer Systems · 2009
The existing clustering algorithm can not discover clusters with arbitrary shape and multi-density using few parameters. In this paper we present a new clustering algorithm named MSTClust which is based on minimum spanning tree. The MSTClust can discover clusters with arbitrary shape and multi-density,can dispose multidimensional data,can detect outer point and have a good expansibility. In allusion to MSTClust we propose an objective function which refers to statistical Information of the weight of edges in minimum spanning tree. Finally the experimental result showed the effectiveness and efficiency of MSTClust and proved that the objective function have good astringency.