150 likes | 257 Vues
This paper introduces innovative techniques for discovering frequent spatiotemporal patterns from the movement data of objects. By employing two algorithms, AIIMOP and MaxMOP, the study aims to assist service providers in delivering targeted information to users in a proactive manner while predicting future user locations. The research integrates a grid-based clustering technique to manage pattern region density and automatically adjust region shapes. The experiments validate the efficiency of these algorithms compared to existing methods, highlighting their potential for optimizing location-based services.
E N D
Manu Shukla Spatiotemporal Pattern Mining Technique for Location-Based Service SystemThi Hong Nhan, Jun Wook Lee and Keun Ho RyuETRI Journal, June 2008
Introduction • Authors propose techniques to discover frequent spatiotemporal patterns from moving objects data • Patterns found can help service provider send information to a user in a push driven manner and predict future location of user • Includes two algorithms AIIMOP and MaxMOP to find frequent and maximal patterns respectively • To control the density of pattern regions and automatically adjust the shape and size of regions, employ grid based clustering technique
Definitions • Trajectory: finite sequence of points {(oj,p1,vt1),(oj,p2,vt2),….,(oj,pn,vtn)} in the XxYxT space where pi is represented by coordinates xi,yi at the sampled time vti for 1<=i<=n • Moving sequence; list of temporally ordered region labels ms=<(a1,t1,(a2,t2),…(aq,tq)> where ai contains oji, ti-ti+1 >>τ and tq-t1 <=max_span.end – max_span.start for q<=T and 1<=i<=q • Subsequence • Frequent Patterns: If ms has support(ms) >= min_sub where min_sub is user-specified, then ms is defined as frequent pattern.
Pattern Movements • Provided function MINE_MOP to allow the adoption of the type of patterns authors wish to obtain with same input • Trajectory reconstructions: results of re-sampling trajectories
Frequent 1-patterns • Decompose a dataset of moving objects into groups of moving points, each Ai={oji|oji ͼ ai} for one timestamp vti • Frequenty 1-patterns are dense regions or clusters discovered from Ai
Frequent k-patterns • Frequent k-pattern is created by merging a pair of frequent 1-patterns in the consideration of the time constraint.
Experiments • Validated efficiency of AIIMOP and MaxMOP under diverse parameters and datasets and by comparing them with grid-based technique using the GSP and DFS_MINE algorithm • Used Synthetic dataset
Conclusions • The patterns mined in algorithms presented can be used to target users • Can be used to make the location-based services more efficient and effective