Efficiently Mining Dynamic Zonal Co-location Patterns Based on Maximal Co-locations

Co-location pattern mining, which discovers feature types that frequently appear together in a nearby geographic region, is an important branch of spatial data mining. With the evolving of computation and communication technology, spatial information is included into more and more datasets. However, existing techniques of mining co-location patterns have to generate all candidate patterns for further examination. The computation cost and requirement of memory space are both high. Therefore, in this paper, we take into account the concepts of maximal pattern to compress the required memory space and to reduce the execution time of mining process. Moreover, we further extend this technique to dynamic zonal co-location pattern mining where co-location patterns in the region dynamically specified by the user will be extracted. Experimental results show that the proposed index structure and mining algorithm can obtain dynamic zonal co-location patterns with high efficiency.