Hybrid recommendation system for tourist spots based on hierarchical sampling statistics and Bayesian personalized ranking
LI Guangli1， ZHU Tao1， HUA Jin1， QIU Diedie2， WU Renzhong2， ZHANG Hongbin2， JI Donghong3
1.School of Information， East China Jiaotong University， Nanchang 330013， China；2.School of Software， East China Jiaotong University， Nanchang 330013， China；3.School of Cyber Science and Engineering， Wuhan University， Wuhan 430072， China
Abstract：Traditional recommendation system based on collaborative filtering only processes the sparse rating matrix. It doesn't extract the deep-level semantics of users (or items) as well as the users' preferences. To alleviate the above issues， a novel recommendation system for tourist spots based on Hierarchical Sampling Statistics (HSS) and Bayesian Personalized Ranking (BPR) is proposed. Users' preferences are generated and described firstly by the HSS algorithm and a subjective evaluation method. Then， deep-level semantics of users (or items) are extracted fully by the Matrix Factorization (MF) algorithm. And the state-of-art BPR algorithm is utilized in turn to optimize the entire recommendation model. Based on the users' preferences and the optimization results of the BPR algorithm， a group of hybrid recommendation results are acquired and supplied to users. We demonstrate the effectiveness of our proposed model via extensive experiments on a novel “smart-travel” dataset created by ourselves. Experimental results show opposed to the best competitor， the RMSE， MAE and F1 value of the presented model improves about 16.59%， 10.05% and 5.04% respectively. Compared against the HSS algorithm， the BPR algorithm has a more prominent role in the recommendation procedure.
2019, Vol. 53 
