Star Graph Neural Networks for Session-based Recommendation

Abstract

Session-based recommendation is a challenging task. Without access to a user's historical user-item interactions, the information available in an ongoing session may be very limited. Previous work on session-based recommendation has considered sequences of items that users have interacted with sequentially. Such item sequences may not fully capture complex transition relationship between items that go beyond inspection order. Thus graph neural network (GNN) based models have been proposed to capture the transition relationship between items. However, GNNs typically propagate information from adjacent items only, thus neglecting information from items without direct connections. Importantly, GNN-based approaches often face serious overfitting problems.We propose Star Graph Neural Networks with Highway Networks (SGNN-HN) for session-based recommendation. The proposed SGNN-HN applies a star graph neural network (SGNN) to model the complex transition relationship between items in an ongoing session. To avoid overfitting, we employ highway networks (HN) to adaptively select embeddings from item representations. Finally, we aggregate the item embeddings generated by the SGNN in an ongoing session to represent a user's final preference for item prediction. Experiments on two public benchmark datasets show that SGNN-HN can outperform state-of-the-art models in terms of P@20 and MRR@20 for session-based recommendation. CCS CONCEPTS• Information systems → Recommender systems.

Hierarchical Context-Aware Recurrent Network for Session-Based Recommendation

“…SGNN-HN [24]: It applies a star graph neural network (SGNN) and a highway network (HN) to model the complex transition relationship between items in a session.…”

“…Qiu et al [23] proposed a weighted graph attention network (WGAT) based approach for generating item representations, which are then aggregated by a Readout function as the user preference. Pan et al [24] apply a star graph neural network (SGNN) and a highway network (HN) to model the complex transition relationship between items in a session.…”

“…e jt = h T j h t (24) where the weight α determines which part of the input sequence should be emphasized or ignored when making predictions. The function e jt specifically computes the energy [29] between the final hidden state h t and the representation of the previous event annotations h j .…”

S-Walk

“…Because sessions are usually sparse by nature, it is useful for alleviating the data sparsity problem by capturing profound relationships among items. (ii) Compared to GNN-based SR models [13,40,41,53,54], it is efficient without requiring complicated hyper-parameter tuning.…”

“…Recurrent neural networks (RNNs) [16][17][18] and attention mechanisms [31,32] have been used to model the sequential dependency of items. Recently, graph neural networks (GNNs) [1,13,40,41,50,53,54] have been used to effectively…”