양방향 인코더와 추가 정보를 이용한 연속적인 POI 추천 모델

초록

Increase Location-Based Social Network Service (LBSNs), POI recommendation has been an active research. Successive POI recommendation, which is used on tourism service and information retrieval etc, recommends next POI based on user’s recent historical check-ins. Existing successive POI recommendation models are unidirectional models, which have limitations when learning user’s mobility pattern. Furthermore, they did not consider additional relevant features sequentially. To address these limitations, we use bidirectional model to represent the context of user’s mobility pattern. Also, integrated various features related to POI(i.e. rating, review, and category)have been considered sequentially. Experiments show that our model has improved performance up to 2.6%.