A fast approach for spatial CO-location pattern mining

Spatial co-location pattern mining means to find subsets of events whose instances are frequently located together in geographic space. Due to the complexity of spatial data types and spatial relationships, co-location mining is much more difficult than traditional transactions mining [1]. With the increasing of the amount of spatial data, computing time it costs grows sharply. In this paper, we present a spatial co-location pattern mining approach called Grid-based method. Then we make use of parallel programming to improve our method. Our Grid-based method would divide a continuous spatial area into many small cells, and each process of our cluster undertakes several cells. Experiment results show that if we just use one process, our method could get the same result as traditional Join-based method [2] while having a faster speed. With the increasing of the number of processes, our algorithm could get a high speed up ratio.