Innovative Research on Deep Learning Algorithm in Intelligent Attractions Recommendation Technology

Abstract

Abstract In order to discover users’ interest preferences from the massive travel information and meet users’ needs for personalized tourist attraction recommendations. In this paper, we construct correspondingly according to the domain knowledge and complete the design and implementation of the mapping according to the corresponding process. The item representation enhancement module is designed to obtain a more comprehensive user preference by extracting information from the user’s historical preferences and then using graph convolution to learn the correlation between the user’s historical behavioral preferences and entities to enhance the KG semantic information of items. An item representation enhancement model based on RippleNet is proposed, from the item representation enhancement module to the interest module, to establish a connection between the user and the implicit information of the item so as to consider their mutual influence and represent the missing information more comprehensively, and finally, a personalized travel recommendation model based on knowledge graph and deep learning is constructed. From the results of the study, it is obtained that the model in this paper is 4.14% and 4.6% better than the best model SASRec in the comparison method in NDCG@10, Recall@10, MRR@10 indicators on both 1.93% higher. At the recommended list length K=20, the Recall rate Recall@20 has reached 80.78%. The superiority of the performance of the model constructed in this paper is demonstrated in the experiments, which brings technical innovation to the personalized attraction recommendation system.