Study Of Spatio-temporal Data Mining Algorithm Based On TAG

Spatio-temporal data mining extracts implicit but potentially valuable information from mass, high dimensional and non-linear spatio-temporal data. Due to both the growth of database archives and the increasing number of spatiotemporal collection equipment, which causes the volume of spatio-temporal data continues to increase significantly automatic and semi-automatic analysis becomes more essential. It is challenging enough for us to find more efficient algorithms to mine useful information from spatio-temporal data.This thesis is to study mining algorithms from large amount of spatio-temporal data. According to the disadvantages of present algorithms in planning evacuation route and mining spatio-temporal co-occurrence patterns, we make research and exploration to improve the computation efficiency and reduce the required memory of algorithm. The main research work is as follows.(1) It improves the calculation efficiency of evacuation route planning algorithm. In order to reduce the number of calculation of the shortest paths in the iteration, we change the weight of edges when the road is paralyzed in evacuation and inform the source node.(2) It defines the problem of mining at most top-K.%MDCOPs without using user-defined thresholds based on time aggregated graph (TAG). We create MDCOP Graph to store the spatio-temporal relationship between instances which improves the computation efficiency of pattern mining and reduces the cost time of algorithm.