Grid-based DBSCAN Algorithm with Referential Parameters

Abstract

A new algorithm GRPDBSCAN (Grid-based DBSCAN Algorithm with Referential Parameters) is proposed in this paper. GRPDBSCAN, which combined the grid partition technique and multi-density based clustering algorithm, has improved its efficiency. On the other hand, because the Eps and Minpts parameters of the DBSCAN algorithm were auto-generated, so they were more objective. Experimental results shown that the new algorithm not only can better differentiate between noises and discovery clusters of arbitrary shapes but also have more robust.