Approaches for finding Optimal Number of Clusters using K-Means and Agglomerative Hierarchical Clustering Techniques

Abstract

Machine learning and pattern recognition both benefit greatly from the study of clustering techniques. Choosing the right number of clusters in cluster analysis is quite tough. Quality of the cluster depends on Optimal number clusters. In this paper, we used three methods such as elbow method, gap statistics method and Silhouette method to find the optimal number of clusters. Furthermore, the agglomerative hierarchical clustering (AHC) algorithm and K-means methods has used for calculating the appropriate number of quality clusters for the data set with the optimal k-value. Both algorithms are evaluated on a given data set using the validation measures such as connectivity, Dunn, and Silhouette to find the optimal number of clusters.