Similarity Measurement And Application Of Indoor Moving-Object Trajectories

A recent report showed that people spend about80percent of their life time in indoor spaces, such as office buildings, shopping malls, airports, and metro-stations. With the development of indoor positioning technologies like RFID, Wi-Fi, and Bluetooth, it is possible for us to obtain locations and trajectories of indoor moving objects. Consequently, how to effectively manage and utilize indoor moving-object data and provide indoor location-based services has been a hot issue in both academia and industries.Based on the background of the development of indoor moving-object data management, in this paper we study similarity measurement of indoor moving-object trajectories and its applications in location recommendation in indoor spaces. Indoor moving-object trajectories often contain rich information about users’ behavior, such as users’interests, location preferences, and moving patterns. Therefore, it has significant and practical values for indoor location-based services to mine and analyze indoor moving-object trajectories.Around the demands of indoor location-based services, this paper mainly focuses on two issues, which are similarity measurement of indoor moving-object trajectories and personalized recommendation of indoor locations. Briefly, we make the following contributions in this paper:(1) We analyze the properties of indoor trajectories and propose a semantic-enhanced similarity-measurement approach for indoor moving-object trajectories. We take into account the characteristics of indoor scenarios and introduce the semantics of locations for measuring trajectory similarity. In addition, we design a data structure called SC_tree to describe the semantic relationships between indoor locations. Accordingly, we define the hierarchical moving patterns of moving objects in indoor spaces, based on which we design a new algorithm to calculate the semantic similarity between trajectories. For spatial similarity measurement, we propose a critical-point-based method to simplify trajectories. Finally, we present the method of computing the similarity between moving-object trajectories by combining semantics and spatial features of indoor trajectories.(2) Based on the similarity measurement of indoor moving-object trajectories, we propose a personalized recommendation algorithm on indoor locations, which employs a clustering step on users’ trajectories and uses the idea of collaborative filtering recommendation. We first propose the user-clustering algorithm based on the similarity measurement of trajectories. Then, we generate the set of candidate locations for recommendation and compute the similarity between users. In addition, we propose a visiting-history-based algorithm to calculate the implicit score between users and locations, which is thereby used to determine users’interests in locations. Finally, we propose a prediction approach to estimate the interest of the query user in location candidates. The location candidates with high rankings according to the prediction scores are selected as the results of recommendation.