Optimum Vdbscan(O-VDBSCAN) For Identifying Downtown Areas

Clustering is an important part of data mining techniques, and VDBSCAN is a well-known density-based one. VDBSCAN is robust against noise and can recognize arbitrary shapes of clusters. Besides, it works effectively when dealing with datasets with varying densities. A main drawback of VDBSCAN is that it still requires a user-specified parameter K. An inappropriate choice of K can seriously degrade the accuracy of results. So in this paper we propose a totally parameter-free algorithm, OVDBSCAN, to find the global optimum K automatically, using the concept of derivative. The basic idea of OVDBSCAN is regarding as the derivative of k-dist, which means the distance between an object and the kth nearest object of it. Then it chooses the largest K on condition that oesn’t excee the threshol we set. In OVDBSCAN, the determination of K is based on the distances among objects within a dataset, thus the generated K reflects the property of this dataset. We’ve applie OVDBSCAN to a two-dimensional sample dataset, and the result shows that it can identify dense areas of varying densities.