Event R-Tree Miner: an Efficient Approach to Mine Sequential Patterns from Spatio-Temporal Event Dataset

Abstract

With the advances of technologies such as GPS, remote sensing, RFID, indoor locating devices, and geosensor networks, mining of spatio-temporal patterns seems more important to track spatio-temporal phenomena with increasingly finer spatial resolutions. Accordingly, we have presented an efficient approach to mine sequential patterns from spatio-temporal event datasets. Here, we make use of three major challenges, such as 1) definition of significance measure for finding spatio-temporal sequential pattern, 2) event tree, to reduce the running time and, 3) the definition of adaptive neighborhood to improve the accuracy. Based on these three challenges, R-tree data structure is utilized to construct the event R-tree for avoiding the accessing of database every time. Furthermore, follow ratio, a significance measure is proposed for finding the significant spatio-temporal sequential patterns and the sequence behavior of spatio-temporal events are identified with the help of adaptive neighborhood that is bounded with MBR condition. Finally, the experimentation is made with the real and synthetic dataset and we have proved the efficiency of the proposed event R-tree miner than the STS miner in terms of computation time