The rapid development of indoor localization techniques such as Wi-Fi and RFID makes it possible to obtain users’ position-tracking data in indoor space. Indoor position-tracking data, also known as indoor moving trajectories, offer many new opportunities to mine decision-making knowledge. In this paper, we study the detection of highly influential positions from indoor position-tracking data, e.g., to detect highly influential positions in a business center, or to detect the hottest shops in a shopping mall according to users’ indoor position-tracking data. We first describe three baseline solutions to this problem, which are count-based, density-based, and duration-based algorithms. Then, motivated by the H-index for evaluating the influence of an author or a journal in academia, we propose a new algorithm called H-Count, which evaluates the influence of an indoor position similarly to the H-index. We further present an improvement of the H-Count by taking a filtering step to remove unqualified position-tracking records. This is based on the observation that many visits to a position such as a gate are meaningless for the detection of influential indoor positions. Finally, we simulate 100 moving objects in a real building deployed with 94 RFID readers over 30 days to generate 223,564 indoor moving trajectories, and conduct experiments to compare our proposed H-Count and H-Count* with three baseline algorithms. The results show that H-Count outperforms all baselines and H-Count* can further improve the F-measure of the H-Count by 113% on average.
Scalable continuous range monitoring of moving objects in symbolic indoor space