Session-based Recommender Systems Using Graph Convolutional Networks

Abstract

The rapid growth of online platforms has led to an overwhelming amount of available information and choices, making personalized recommendations crucial for enhancing user experience and satisfaction. In this thesis, we focus on session-based recommender systems, which aim to provide accurate recommendations by considering users' sequential behaviour and short-term interests. To address the challenges posed by session-based recommendations, we leverage the power of Graph Convolutional Networks (GCNs). GCNs have shown remarkable effectiveness in modelling complex relationships and capturing the underlying structure of data. By exploiting the graph-like nature of user sessions, we harness the potential of GCNs to capture the intricate dependencies between items and uncover latent patterns within sessions