Mining Mixed-drove Co-occurrence Patterns For Large Spatio-temporal Data Sets: A Summary of Results

摘要

Discovering mixed-drove spatiotemporal cooccurrence patterns (MDCOPs) is an important field with many applications such as identifying tactics in battlefields, crime detection, etc. In practical applications, it is difficult to mine MDCOPs from large spatio-temporal data sets. Firstly, mining MDCOPs is computationally very expensive because the set of candidate co-occurrence instances is exponential in the number of object-types. Secondly, the spatio-temporal data sets are large and can't be managed in memory. In order to reduce the number of candidate co-occurrence instances, we present a novel and computationally efficient MDCOP Graph Miner algorithm by using Time Aggregated Graph. The LDMDCOP Graph Miner algorithm is presented, which can deal with large data sets by means of file index. The correctness, completeness and efficiency of the proposed methods are analyzed. Experimental results show that the proposed MDCOP Graph Miner is computationally more efficient than the fast MDCOP-Miner and the LDMDCOP Graph Miner can effectively deal with the large spatiotemporal data sets.