Unsupervised Graph Embeddings for Session-based Recommendation with Item Features
Andreas Peintner, Marta Moscati, Emilia Parada-Cabaleiro, Markus Schedl, Eva Zangerle
TL;DR
This paper tackles the challenge of improving session-based recommendations by jointly leveraging item features and graph structure. It introduces GCNext, a two-stage, unsupervised framework that builds a feature-rich item co-occurrence graph and learns embeddings via Bootstrapped Graph Latents with a GATv2 encoder, which are then used to initialize downstream sequential models and refine nearest-neighbor methods. The approach yields consistent gains across a range of base models and datasets, with notable improvements in $MRR@20$ (up to 12.79%) and faster convergence when using pre-trained embeddings. The work demonstrates that modularly integrating graph-based item representations can significantly enhance the performance of state-of-the-art session-based recommender systems and suggests directions for future work in cold-start scenarios and broader graph-embedding comparisons.
Abstract
In session-based recommender systems, predictions are based on the user's preceding behavior in the session. State-of-the-art sequential recommendation algorithms either use graph neural networks to model sessions in a graph or leverage the similarity of sessions by exploiting item features. In this paper, we combine these two approaches and propose a novel method, Graph Convolutional Network Extension (GCNext), which incorporates item features directly into the graph representation via graph convolutional networks. GCNext creates a feature-rich item co-occurrence graph and learns the corresponding item embeddings in an unsupervised manner. We show on three datasets that integrating GCNext into sequential recommendation algorithms significantly boosts the performance of nearest-neighbor methods as well as neural network models. Our flexible extension is easy to incorporate in state-of-the-art methods and increases the MRR@20 by up to 12.79%.