DDBSCAN: Different Densities-Based Spatial Clustering of Applications with Noise

Recent advances in using computer with different fields of sciences produced huge amounts of data. These data represent as an analysis tool and key to overcome many problems. Clustering is a primary process to analyze the data as well as, it´s a preprocessing step before other techniques like classification. Density-Based clustering algorithms have advantages like clustering any arbitrary shapes and defining number of clusters according to database. DBSCAN (Density Based Spatial Clustering of Application with Noise) [1] is the basic density-based algorithm. But it fails to discover different densities clusters, adjacent clusters and finally some noise points among different densities clusters. This paper addresses DBSCAN problems and tries to solve these problems by developing DDBSCAN. The basic idea is to compute the density of a cluster with respect to radius value Eps and minimum number of points MinPts. Then provide density threshold which is the responsible for joining a point to a certain cluster or not. Experiments show that DDBSCAN outperforms DBSCAN in different densities and adjacent clusters datasets.