Discovery of useful patterns from human movement behavior can convey valuable knowledge to a variety of critical applications. Existing approaches focus on outdoor group discovery and mainly consider objects who belong to the same cluster as a possible group, which leads to the inability to discover all the existing groups. This is especially true for indoor human-generated trajectories, where spatially distant objects could be related to one group. Considering the human movement characteristic, we propose the loose travelling companion pattern which allows objects in different clusters to form a group, as long as the community of clusters doesn't change during the movement and all members stay together in the limited number of times. To tolerate the unrealistic temporary clusters, we extend the algorithm to the weakly consistent travelling companion pattern which relaxes the continuous requirement. In this paper, we also introduce a smart trolley which is used to collect the passenger movement data at airports in order to discover the groups. The acquired knowledge will then be applied to provide personalized services and advertisement. By the experimental analysis and comparison with the real and synthetic datasets, it is shown that the proposed approach can discover more complete and accurate groups.
|Title of host publication||2016 IEEE 18th International Conference on High Performance Computing and Communications; IEEE 14th International Conference on Smart City; IEEE 2nd International Conference on Data Science and Systems (HPCC/SmartCity/DSS)|
|Number of pages||8|
|Publication status||Published - 14 Dec 2016|
- human trajectories
- movement pattern
- spatio-temporal data mining
- group relationship