Mining patterns from Spatio-Temporal Data has been an attractive area of recent research for its wide applications. Some approaches are extended to Semantics Rich Trajectory Data of Moving Objects. Semantics contain contextual information such as health or behavioral information, location description, etc. In Frequent Itemset Mining, micro analysis has been done, to capture significant time intervals in which itemsets have more support than global support, to know the deeper knowledge such as huge demand for goods or peak surge in virus spread.Aim of this paperis to find out maximal significant hot spots of frequent semantic patterns indicating the cluster of locations at which the patterns occur most frequently in chunks of time-stamps. We propose SPT Tree to find local and then global frequent semantic patterns valid on a Spatio-Temporal Semantic Trajectory data set and then we identify significant time intervals for frequent semantic patterns. Finally, significant hot spots are identified based on space proximity among the locations in maximal significant time interval. To find space proximity, we follow a two-step process. First, we find initial clusters of locations using k-Means algorithm with Manhattan metric and next we train the supervised learning based Mahalanobis distance metric so that our approach similar to DBScan Clustering algorithm encloses the locations along the dimensions of natural spread of the trajectory points. Experimental results are presented.
Volume 12 | Issue 6
Pages: 1488-1502
DOI: 10.5373/JARDCS/V12I2/S20201346