A Parallel Spatial Co-location Mining Algorithm Based on MapReduce

Spatial association rule mining is a useful tool for discovering correlations and interesting relationships among spatial objects. Co-locations, or sets of spatial events which are frequently observed together in close proximity, are particularly useful for discovering their spatial dependencies. Although a number of spatial co-location mining algorithms have been developed, the computation of co-location pattern discovery remains prohibitively expensive with large data size and dense neighborhoods. We propose to leverage the power of parallel processing, in particular, the MapReduce framework to achieve higher spatial mining processing efficiency. MapReduce-like systems have been proven to be an efficient framework for large-scale data processing on clusters of commodity machines, and for big data analysis for many applications. The proposed parallel co-location mining algorithm was developed on MapReduce. The experimental result of the developed algorithm shows scalability in computational performance.