Hausdorff distance and global silhouette index as novel measures for estimating quality of biclusters

Biclustering is a commonly used technique for extracting local patterns from microarray data, for which several algorithms have been proposed. Hence it is important to define metrics that compare the various algorithms. In this paper, we have defined novel measures of hausdorff distance between biclusters and global silhouette index for estimating the quality of biclusters extracted by the existing algorithms. We have also compared these measures with the standard measures such as the proportion of enriched biclusters for benchmark biological datasets. Our experimental results show almost similar variation of all these metrics for most of the datasets. The computation of these metrics for a given dataset for all the existing algorithms gives the most suited algorithm for the considered dataset.