New approach for finding number of clusters using distance based k-means algorithm

Clustering is one of the most universal unsupervised classification methods for partitioning objects into a set of meaningful clusters. The k-means clustering algorithm is a commonly used partitioning based clustering method for finding optimal number of clusters. However, number of clusters generated by k-means algorithm depends on the choice of centroid value which sometimes could be misled. Therefore, a new approach for identifying the optimal number of clusters based on distance in k-means algorithm is proposed. The designed algorithm was tested using twelve sets of simulated data has revealed that the proposed algorithm is able to identify the exact number of clusters.