This project aims to delve into the intricacies of recommendation systems, with a primary focus on understanding and addressing the cold start problem and the challenges associated with modeling complex relationships between items and users. The motivation behind this exploration is to implement the Contextualized Graph Attention Network (CGAT Network) from the paper in citation, which introduces a novel approach to enhance recommendation accuracy in the face of limited user and item data.
Key Features Collaborative Filtering: The first version of our recommendation system employs collaborative filtering techniques to provide personalized recommendations based on user behavior and preferences.
Contextualized Graph Attention Network (CGA-Net): The second version of our recommendation system introduces the CGA-Net approach, a GNN model that considers item features, context information, and utilizes a graph attention mechanism to make more accurate and context-aware recommendations.
@article{liu2021contextualized,
Title={Contextualized Graph Attention Network for Recommendation with Item Knowledge Graph},
Author={Liu, Yong and Yang, Susen and Xu, Yonghui and Miao, Chunyan and Wu, Min and Zhang, Juyong},
Journal={IEEE Transactions on Knowledge and Data Engineering},
Year={2021},
Publisher={IEEE}}