Abstract—Personalized restaurant recommendation is a hotpot in the field of smart tourism. Most existing methods focus on user preference mining based on user portraits but ignore the importance of mining feature based on integrating multi-source information. Hence, a personalized restaurant recommendation method based on multi-attribute mining (MAM_ResR) is proposed. Having constructed restaurant knowledge graph, combining geospatial semantics obtained from dining sequence trajectory and inherent attribute semantics mined by network embedding of restaurant-attribute graph extracted from restaurant knowledge graph, features of restaurants fusing multi-source information are obtained. Then, using explicit feedback and dining trajectory, the user preference fusing multi-attribute feature is obtained. Finally, the recommendation list is calculated by similarity between restaurant characteristics and user preferences. Experiments based on real data show that it is very effective to integrate multiple types of information into the recommendation method to improve recommendation performance.
Index Terms—Multi-attribute mining, sequence representation learning, network representation learning, explicit feedback, personalized restaurant recommendation.
The authors are with the Guilin University of Electronic Technology, Guilin 541004, China (e-mail: 925211089@qq.com, changl@guet.edu.cn).
[PDF]
Cite: Kuang Haili, Changliang, Sun Yanpeng, and Bin Chenzhong, "A Multi-attribute Mining Based Personalized Restaurant Recommendation Method," International Journal of Knowledge Engineering vol. 5, no. 1, pp. 8-14, 2019.
