Ships' Spatiotemporal Co-Occurrence Patterns Mining Based on Sliding Spatiotemporal Cuboid and Improved Support

There exists two problems when using traditional methods to mine the spatiotemporal co-occurrence pattern of marine vessels: First, the traditional methods based on spatial participation index Pi and temporal participation index TPi may not be able to find co-occurrence patterns across time slots; Second, when the amount of AIS data is large, traversing the distance and time interval between every two instances will seriously increase the time consuming of the algorithm. Third, some ships in some patterns left the few AIS record, so there are few transactions that include this pattern. Although the co-occurrence characteristics of ships in the pattern are very obvious, the pattern will also be missed by traditional discrimination method based on Pi and TPi. Some ships in some patterns left the more AIS record, so there are more transactions that include the pattern. Although the co-occurrence characteristics of ships in the pattern are not significant, the pattern of low value will still be judged as spatiotemporal co-occurrence pattern. To solve these problems, the boats' spatiotemporal co-occurrence patterns mining method based on Sliding Spatiotemporal Cuboid and Improved Support (SSC-IS Method) is presented. On the one hand, the improved support which can adaptively adjust the calculation expression of the support according to the pattern of different ships is proposed to revise the discrimination criterion; on the other hand, SSC-IS Method uses Spatiotemporal Cuboid structure to store AIS data and introduces the new operation named split plane random slide to obtain as many co-occurrence patterns which across time slots as possible. The real AIS data is used to test the performance of the algorithm and the result shows that compared with the traditional method, SSC-IS Method can handle a large amount of data that is very difficult to handle by traditional method and obtain transactional databases on the basis of saving computing time. Furthermore, under the same constraint of the threshold, the SSC-IS method can mine the more spatiotemporal co-occurrence patterns.