Scenic Spot Recommendation Method Integrating Knowledge Graph and Distance Cost

Yue Shen, Xiaoxu Zhu

Abstract: Aiming at the problem that the traditional collaborative filtering algorithm only considers external ratings and cold start when recommending attractions, this paper proposes an attraction recommendation algorithm integrated with knowledge graph. Firstly, we construct a user rating matrix based on user ratings and the number of reviews, and calculate the similarity of attractions. Then, we use TransR model to train the semantic vector matrix of attractions, and use cosine similarity formula to calculate the semantic similarity of attractions. Finally, the two similarities are fused and applied to ALS matrix factorization. At the same time, in order to make the model pay attention to user preferences and take into account the distance elements in the scenic spots for recommendation, the distance cost is integrated in the loss function. Experimental results on the scenic spots dataset show that the proposed algorithm is better than the traditional method in scenic spots recommendation.