Abstract:

Clustering is the assignment of a set of observations into subsets so that observations in the same cluster are similar in some sense, observations in the different cluster are different. An improved clustering method based on MST is raised out. This method can determine discrepancy more exactly, which is more consistent with vision sensing process of human being. Evaluation of clustering shows that it can increase effectiveness of cluster. This method can be used in the classification and recognition of information.

Key words: minimum spanning tree, discrepancy, cluster, domain of influence, effectiveness