Attention-based recurrent neural network for location recommendation

Due to the rapid development of Location-Based Social Networks (LBSNs), the Point of Interest (POI) recom­mendation has been attracted a lot of research attention. Based on the LBSNs, users are able to share their relevant visiting experience via check-in records. The sequential check-in data not only explicitly show users' moving trajectories, but also implicitly describe personal preferences and corresponding life patterns based on different contexts (e.g., time and geographical locations). The traditional POI recommender systems only consider common contexts (e.g., visit frequency, distance, and social relationship), but ignore the significance of life patterns for individuals during different time periods. In addition, current recommender systems hardly provide interpretable and explainable recommendations based on these limited contexts. In this paper, we propose an Attention-based Recurrent Neural Network (ARNN) to provide an explainable recommendation based on the sequential check-in data of the correspond­ing user. Our proposed approach makes use of the sequential check-in data to capture users' life pattern and utilizes a deep neural network to provide transparent recommendations. The major contribution of this paper are: (1) the proposed model is capable of providing explainable recommendations based on life patterns which implicitly describes the preference of the corresponding user; (2) the proposed approach is able to design a visiting plan (i.e., a series of recommendations) based on users' past visiting patterns instead of simply showing top-N recommendations; (3) we evaluate our proposed approach against a real world dataset and compare it to other start-of-the-art approaches.