| Summary: | Numerous clustering algorithms become invalid when data classes are unbalanced and close to each other, or clusters are arbitrary shapes and have different densities. A novel region density based clustering algorithm, DEDIC, is proposed to meet this challenge. It is based on recognizing dense regions using mutual nearest neighbors and estimating region density of a cluster by computing maximum of the directly-reachable distances among core points within the cluster. DEDIC is superior to the popular density-based clustering algorithm DBSCAN in two aspects. First, since there is only one parameter (choice of k nearest neighbors), the algorithm complexity is reduced, and second, an improved ability to handle clusters with large density variations. The superiority of DEDIC is demonstrated on several artificial and real-world datasets with respect to nine known clustering approaches.
|
|---|