Spatial Clustering Algorithm Based on Optimized-Division

Traditional grid-density based spatial clustering algorithms divide input data space into partitions with same width and neglect the natural distributing character of initial data space. A new robust spatial clustering algorithm based on optimized-division (OpD-Clus) is proposed in this paper. Spatial data space is divided by hyper-planes which are encertained with axis-paralleled histogram in OpD-Clus algorithm. Division of data space relies on natural distributing character of input data space to improve the accuracy and efficiency of spatial clustering. Simultaneously, the outstanding difference between density-region and spare-region makes setting of density threshold parameter easily and reduces the parameter dependence of spatial clustering algorithm. The validity, efficiency and un-sensitivity of parameters of OpD-Clus algorithm is demonstrated by experiment results.