Att-KGCN: Tourist Attractions Recommendation System by using Attention mechanism and Knowledge Graph Convolution Network
This work addresses a domain-specific problem for tourism recommendation systems, but it is incremental as it builds on existing knowledge graph-based methods.
The paper tackled the problem of selecting suitable tourist attraction attributes for recommendations in tourism by proposing Att-KGCN, an improved model that uses attention and knowledge graph convolution to automatically discover semantically similar neighboring entities, resulting in good recommendation effects as verified through experiments on a dataset from Socotra Island-Yemen.
The recommendation algorithm based on knowledge graphs is at a relatively mature stage. However, there are still some problems in the recommendation of specific areas. For example, in the tourism field, selecting suitable tourist attraction attributes process is complicated as the recommendation basis for tourist attractions. In this paper, we propose the improved Attention Knowledge Graph Convolution Network model, named ($Att-KGCN$), which automatically discovers the neighboring entities of the target scenic spot semantically. The attention layer aggregates relatively similar locations and represents them with an adjacent vector. Then, according to the tourist's preferred choices, the model predicts the probability of similar spots as a recommendation system. A knowledge graph dataset of tourist attractions used based on tourism data on Socotra Island-Yemen. Through experiments, it is verified that the Attention Knowledge Graph Convolution Network has a good effect on the recommendation of tourist attractions and can make more recommendations for tourists' choices.