Smart tourism services and resource optimisation based on big data and knowledge graphs

Abstract

To address the challenges of information overload and resource misallocation in the tourism industry, this paper proposes an intelligent service framework that integrates multi-source big data with knowledge graphs. By constructing a tourism-specific knowledge graph from the Yelp dataset (containing over 12,537 POIs and 45,821 users) and combining relational graph convolutional networks with long short-term memory models, the framework achieves precise personalised recommendations and dynamic resource optimisation. The proposed multi-task learning architecture jointly optimises recommendation accuracy and resource prediction performance. Extensive experiments show that the model significantly outperforms baseline methods, achieving a Precision@10 of 0.0914 and Recall@20 of 0.2542, along with a 21.73 root mean square error in flow prediction – demonstrating notable improvements in interpretability and robustness. This study provides an effective technical pathway for enhancing tourism service intelligence and operational efficiency.