Abstract: Data clustering is an unsupervised task that can generate different shapes of clusters for a particular type of data set. Hence choosing an algorithm for a particular type of data set is a difficult problem. This study presents the choice of an appropriate clustering algorithm by a comparative study of three representative techniques like K-means, Kohonen`s Self Organizing Map (SOM) and Density Based Spatial Clustering of Applications with Noise (DBSCAN) based on the extensive simulation studies. Comparison is performed on the basis of cluster quality index `ß`, percentage of samples correctly classified and CPU time. The experimental results show that if the clusters are of arbitrary shape, a density based clustering algorithm like DBSCAN is preferable, where as if the clusters are of hyper spherical or convex shape and well-separated then the SOM or K-means is preferable.