A Comparison of Validation Indices for Evaluation of Clustering Results of DNA Microarray Data

Abstract

In this paper we assess the clustering results of gene expression (microarray) data by using different validation indices. We perform k-means clustering and evaluate the results by using three validation indices: Silhouette, Davies-Bouldin's and our proposed BR index. For performing the experiments we use our implementation of these measures in MATLAB. From the comparison of the validation indices we can conclude that BR index has better decreasing tendency depending on the number of clusters than Davies Bouldin's index for yeast microarray dataset.