A Clustering Method Based on the Modified RS Validity Index

摘要

This paper describes a new method to the determination of the optimal number of well-separable clusters in data sets. The determination of this parameter is necessary for many clustering algorithms to define the naturally existing clusters correctly. In the presented method the idea of the agglomerative hierarchical clustering has been used, and the modified RS cluster validity index has been applied. In the first phase of the method, clusters are created due to the idea of hierarchical clustering. Then, for the optimal number of clusters the k-means algorithm is performed. The method has been used for multidimensional data, and the received results confirm very good performances of the proposed method.