DBRS: A Density-Based Spatial Clustering Method with Random Sampling

When analyzing spatial databases or other datasets with spatial attributes, one frequently wants to cluster the data according to spatial attributes. In this paper, we describe a novel density-based spatial clustering method called DBRS. The algorithm can identify clusters of widely varying shapes, clusters of varying densities, clusters which depend on non-spatial attributes, and approximate clusters in very large databases. DBRS achieves these results by repeatedly picking an unclassified point at random and examining its neighborhood. If the neighborhood is sparsely populated or the purity of the points in the neighborhood is too low, the point is classified as noise. Otherwise, if any point in the neighborhood is part of a known cluster, this neighborhood is joined to that cluster. If neither of these two possibilities applies, a new cluster is begun with this neighborhood. DBRS scales well on dense clusters. A heuristic is proposed for approximate clustering in very large databases. With this heuristic, the run time can be significantly reduced by assuming that a probabilistically controlled number of points are noise. A theoretical comparison of DBRS and DBSCAN, a well-known density-based algorithm, is given. Finally, DBRS is empirically compared with DBSCAN, CLARANS, and k-means on synthetic and real data sets.