Finding top-k similar users based on Trajectory-Pattern model for personalized service recommendation

Service recommendation for Location-based Social Networks (LBSNs) has been received much attention mainly in the field of large-scale Wireless Communication Networks. It is a promising solution to guarantee service reliability and improve quality of experience (QoE) of users. Many schemes have been proposed to provide the reliable and personalized services for users, such as collaborative filtering (CF), content-based recommendation(CBR), and context-aware recommendation, etc. However, the activity sequence and type popularity is lacked in above researches, which leads to the recommendation accuracy and service experience decline. In this paper, we propose TopSim-TP, a novel trajectory-pattern-based similar users discovery method aiming to effectively mine the life cycle and preference of users. First, we describe the mobile behavior of users by mapping location information into type information. Then, we model the trajectory-pattern for each user. Finally, we present a Common Trajectory-Pattern (CTP) extraction algorithm and compute the similarity of sequences to find the top-k similar users for current user. The experimental results show that our method outperforms the existing method.