A Location-Optimized Clustering Algorithm Based on Hierarchies and Density

A new kind of clustering algorithm called LOCAHID is presented in this paper. LOCAHID views each potential cluster as a tight coupling structure, which can be described by a density tree. Every density tree is dynamically generated according to its local density distribution. Those "closer" clusters are merged if some conditions are satisfied. In order to extend its applications to large data sets, a typical location-optimized technology is introduced to lower its running time and space storages. LOCAHID inherits the strongpoints of hierarchical methods and density-based methods, such as preferable accuracy in discovering clusters with arbitrary shape, good ability of processing noise data sets, weak sensitivity to input parameters and no limitation of global density threshold. The experiments illustrate the effectiveness