Density Peaks Spatial Clustering by Grid Neighborhood Search

In the application of spatial data clustering, the density-based clustering method can achieve good results. DPC algorithm is a density-based clustering algorithm, which can discover the clustering of irregular shapes. The algorithm is trustworthy of clustering results, simple to implement, and parameter robust. However, the DPC algorithm needs to calculate the distance between the two pairs. It takes a long time to calculate the local density and high-density distance for large-scale spatial data sets. To solve the problem of low efficiency in large datasets, this paper improved the DPC algorithm and proposed a density peak clustering algorithm, DPSCGNS, based on grid neighborhood search. DPSCGNS map raw data to grid cells and redefine the local distance and high-density distance of grid cells. By using the grid to index neighborhood information, the local density and high-density distance of grid cells can be calculated rapidly. Experiments on several data sets demonstrate that the efficiency of DPSCGNS algorithm is improved without decline on clustering effect.