Comparative study of cluster validity techniques using K-mediod algorithm

The most important task of clustering process is the validation of results obtained from clustering algorithms. There are many cluster validation criteria's but the most commonly used approaches are founded on internal validity indices. There are numerous indices that have been suggested from time to time but there are only some of them that have been popularly used. In this paper we have drawn a comparative analysis of external and internal validity measures using clustering results from K-mediod algorithm; we show the results of our experimental work which can be useful in selecting the most suitable index and providing an insight about the performance of different indices on different datasets. We have used four datasets: Iris, Gene dataset, liver disorder and Seeds datasets from UCI repository in our experiment.