English

Abstract

Determining an optimal number of clusters and producing reliable results are two challenging and critical tasks in cluster analysis. We propose a clustering method which produces valid results while automatically determining an optimal number of clusters. Our method achieves these results without user input pertaining directly to a number of clusters. The method consists of two main components: splitting and merging. In the splitting phase, a divisive hierarchical clustering method (based on the DIANA algorithm) is executed and interrupted by a heuristic function once the partial result is considered to be “adequate”. This partial result, which is likely to have too many clusters, is then fed into the merging method which merges clusters until the i¬nal optimal result is reached. Our method’s effectiveness in clustering various data sets is demonstrated, including its ability to produce valid results on data sets presenting nested or interlocking shapes. The method is compared with cluster validity analysis to other methods to which a known optimal number of clusters is provided and to other automatic clustering methods. Depending on the particularities of the data set used, our method has produced results which are roughly equivalent or better than those of the compared methods.