Detecting Hotspots from Trajectory Data in Indoor Spaces

Abstract

The increasing deployment of indoor positioning technologies like RFID, Wi-fi, and Bluetooth offers the possibility to obtain users’ trajectories in indoor spaces. In this paper, based on indoor moving-object trajectories, we aim to detect hotspots from indoor trajectory data. Such information is helpful for users to understand the surrounding locations as well as to enable indoor trajectory mining and location recommendation. We first define a new kind of query called indoor hotspot query. Then, we introduce a pre-processing step to remove meaningless locations and obtain indoor stay trajectories. Further, we propose a new approach to answering indoor hotspot queries w.r.t. two factors: (1) users’ interests in indoor locations, and (2) the mutual reinforcement relationship between users and indoor locations. Particularly, we construct a user-location matrix and use an iteration-based technique to compute the hotness of indoor locations. We evaluate our proposal on 223,564 indoor tracking records simulating 100 users’ movements over a period of one month in a six-floor building. The results in terms of MAP, P@n, and nDCG show that our proposal outperforms baseline methods like rank-by-visit, rank-by-density, and rank-by-duration.