An Extended DBSCAN Clustering Algorithm

Finding clusters of different densities is a challenging task. DBSCAN “Density-Based Spatial Clustering Applications with Noise” method has trouble discovering various since it uses fixed radius. This article proposes an extended for finding densities. The proposed dynamic radius and assigns regional density value each object, then counts the objects similar within If neighborhood size ≥ MinPts, object core, cluster can grow from it, otherwise, assigned noise temporarily. Two are in local if their similarity threshold. discover any data effectively. requires three parameters; Eps (distance to kth neighbor), practical results show superior ability suggested detect even no discernible separations between them.