Jinduk Park, Yonsei University
We often consider various rating criteria such as cleanliness, price, and location when booking a hotel. Such rating information based on specific criteria significantly influences consumer decisions, rather than just single ratings. On the other hand, Graph Neural Networks (GNNs) become a state-of-the-art method in recommendation systems. Despite this, there is currently no methodology that utilizes multi-criteria rating information with GNNs. To effectively utilize multi-criteria rating information and enhance recommendation performance, however, a specially designed methodology considering multi-criteria rating environments is required. In this study, we propose a novel GNN called CPA-LGC, which utilizes multi-criteria rating information to construct a graph structure and learn the criterion preferences of each user via GNN. Experience and leverage the pioneering work of GNN methodology in multi-criteria recommendation using CPA-LGC to achieve superior recommendation performance!
