Graph Neural Networks for Recommendation Systems

• Github URL: Project Link

Used Graph Neural Networks (GNNs) to create a recommendation system and learn the joint embeddings of each user and item which are part of the given graph.

Trained the model to predict the rating of an item by a user by utilizing information from two graphs - user-user social graph and user-item graph.

Carried out experiments on two real-world datasets - Ciao and Epinions dataset and obtained mean absolute error of 0.71 and 1.04 respectively.