Top-K Item Recommendations for Content Curation Platforms Using a Graph Convolutional Autoencoder

Abstract

There are millions of users using applications where content creators (curators) create items or content for the users to consume. With users spending more and more time on these platforms, personalized recommendations incentivizes users to consume more content. Much research has been done to improve the performance of the systems but the process of using a graph-based relationship to map the users and curators is still a largely new topic that has the potential to capture the relationships between users, content, and curators. In this thesis, we propose a graph- based convolutional autoencoder recommender system for top-K item recommendations for each user. We compare the results of our model to current state of the art recommender systems and offer insight into the noise that impacts the relationship between users and curators. We demonstrate that our model performs similarly to current state of the art models and provide future directions that require more research.