A Multi-purpose Density Based Clustering Framework

Abstract

In this paper, we present a multi-purpose density-based clustering framework. The framework is based on a novel cluster merging algorithm which can efficiently merge two sets of DBSCAN clusters using the concept of intersection points. It is necessary and sufficient to process just the intersection points to merge clusters correctly. The framework allows for clustering data incrementally, parallelizing the DBSCAN algorithm for clustering large data sets and can be extended for clustering streaming data. The framework allows us to see the clustering patterns of the new data points separately. Results presented in the paper establish the efficiency of the proposed incremental clustering algorithm in comparison to IncrementalDBSCAN algorithm. Our incremental algorithm is capable of adding points in bulk, whereas IncrementalDBSCAN adds points, one at a time.