Research on Travel Accommodation Recommendation Algorithm Based on Multimodal Neural Network

Most of the existing travel accommodation recommendation methods only consider a single mode, but the dependence of a single information source is too strong to fully capture the complex relationship between multimodal information and the computational efficiency is low. Based on this, we propose a travel accommodation recommendation algorithm based on multi-modal heterogeneous graph neural network. A variety of modal features of travel and accommodation information are obtained by using Vision Transformer (ViT) and Bidirectional Encoder Representations from Transformer (BERT) models. Firstly, it is too specific to fine-tune the part of Token in the Block layer of the Encoder part of the model, and the various modal messages are combined with the graph neural network to improve the semantic expression of nodes in the graph neural network by using the rich semantic information in the multimodal information. Then the information is constructed into a heterogeneous graph with more types of nodes and edges to fully mine high-order information to alleviate common problems such as matrix sparse and cold start problems. Finally, comparative experiments and ablation experiments are carried out on THUCNews data sets, and the results show that the accuracy of this method is improved compared with the traditional graph neural network recommendation algorithm.