Natural phenomena show that many creatures form large social groups and move in regular patterns. However, existing object tracking applications focus on finding the movement patterns of individual objects or all objects. Therefore, we study the group movement pattern mining problem of identifying objects with similar movement patterns based on their movement trajectories. The discovered information is important in some biological research domains, such as the study of animals’ social behavior and wildlife migration. It also has wide applications, such as object behavior prediction, tracking network design and data management, traffic information query for intelligent vehicle design, social relationships for information sharing or advertisement, etc.
To solve the group movement mining problem, we proposed a distributed framework comprising a local mining phase and a cluster ensembling phase. In the local mining phase, we model and find movement patterns hidden in local trajectory dataset. Then, based on the derived patterns, we proposed a new similarity measure to compute the similarity of moving objects and identify the local group relationships. To address the energy conservation issue in resource-constrained environments like WSNs, only the local grouping results are transmitted to a central node for further ensembling. In the cluster ensembling phase, we combine the local grouping results to remove inconsistency and improve the grouping quality by using the information theory. The distributed framework provides efficiency for knowledge discovery as well as flexibility to adapt to various local settings.
In this talk, I would like to introduce our preliminary study results and share our current research on trajectory data mining and it application.