A Hierarchical Approach for Clusters in Different Densities

Clustering has the following challenges: 1) clusters with arbitrary shapes; 2) minimal domain knowledge to determine the input parameters; 3) scalability for large data sets. Density-based clustering has been recognized as a powerful approach for discovering clusters with arbitrary shapes. However, the other two challenges still remain in most existing clustering algorithms. In this paper, we explore a hierarchical and iterative densitybased clustering method for large data sets with clusters in different densities. We meet the second challenge by reducing input parameters and solve the third challenge by means of hashing techniques and a vertical data structure, P-tree1 . Our experiments with three different data sets show that our approach is more efficient and robust than DBSCAN, TURN*, and K-means with better clustering qualities.