Crowdsensing Service by Mining Trajectory Patterns in Wireless Sensor Networks

With the advances of wireless technologies, moving objects or users equipped with sensors are located and tracked by wireless sensor networks. Recommendations for the direction and route in exhibition by analyzing and mining the user's movement behavior can be conducted. In this paper, we study how to mine the data of historical trajectories and develop an approach which can guide the users. When users want to move out to the other region, the users are guided to move to the suggested point of interest by the corresponding sensor with the guided direction. After that, the moving user is attempted to be guided to move in the new region covered by a large size area with high attractive importance. According to the historical trajectories for users, three approaches are proposed to find out the importance for each sensor node such that the guided direction and the size of covered region can be computed dynamically. Through the computed joint forces, the points of interest in the space can be partitioned to many regions with different sizes in terms of Voronoi diagram techniques. Finally, the experimental results show that the moving user can be guided to the proper point of interest based on our proposed approaches with the considerations of visited frequency, dwell time and user movement speed.