Abstract:The problem of session-aware recommendation aims to predict users' next click based on their current session and historical sessions. Existing session-aware recommendation methods have defects in capturing complex item transition relationships. Other than that, most of them fail to explicitly distinguish the effects of different historical sessions on the current session. To this end, we propose a novel method, named Personalized Graph Neural Networks with Attention Mechanism (A-PGNN) for brevity. A-PGNN mainly consists of two components: one is Personalized Graph Neural Network (PGNN), which is used to extract the personalized structural information in each user behavior graph, compared with the traditional Graph Neural Network (GNN) model, which considers the role of the user when the node embedding is updated. The other is Dot-Product Attention mechanism, which draws on the Transformer net to explicitly model the effect of historical sessions on the current session. Extensive experiments conducted on two real-world data sets show that A-PGNN evidently outperforms the state-of-the-art personalized session-aware recommendation methods.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is, in session - aware recommendation, the deficiencies of existing methods in capturing complex item - transition relationships and the flaws of failing to clearly distinguish the influence of different historical sessions on the current session. Specifically: 1. **Capturing complex item - transition relationships**: Existing session - aware recommendation methods are difficult to capture the complex transition relationships between items, especially between different sessions. For example, a user may show different points of interest in different sessions, and the transition relationships between these points of interest are very complex, while traditional sequence models are difficult to capture this complexity. 2. **Distinguishing the influence of historical sessions**: Most existing session - aware recommendation methods do not clearly distinguish the influence of different historical sessions on the current session. For example, if a user has previously clicked on a digital camera, and currently clicks on SD cards and Micro - SD cards, then there is a strong association between the historical session and the current session. But if the current click is on automotive products, then the association between the historical session and the current session is weak. However, existing models often ignore this fine - grained influence, resulting in poor recommendation effects. To solve these problems, the paper proposes a new method - Attention - based Personalized Graph Neural Network (A - PGNN). A - PGNN mainly consists of two components: 1. **Personalized Graph Neural Network (PGNN)**: It is used to extract personalized structural information in each user behavior graph. Compared with the traditional Graph Neural Network (GNN) model, PGNN takes into account the user's role when updating node embeddings. 2. **Dot - Product Attention**: Drawing on the Transformer network, it is used to explicitly model the influence of historical sessions on the current session. Through these two components, A - PGNN can more effectively capture the complex transition relationships between items and can clearly distinguish the influence of different historical sessions on the current session, thereby improving the performance of the recommendation system. Experimental results show that A - PGNN is significantly superior to existing personalized session - aware recommendation methods.

Personalized Graph Neural Networks with Attention Mechanism for Session-Aware Recommendation

Feature-Level Attentive Neural Model for Session-Based Recommendation

Personalized Session-Based Recommendation Using Graph Attention Networks.

Category-aware Self-Supervised Graph Neural Network for Session-Based Recommendation

A Spatiotemporal Graph Neural Network for session-based recommendation

Attentive Sequential Model Based on Graph Neural Network for Next Poi Recommendation.

High-order attentive graph neural network for session-based recommendation

Session-Based Recommendation with Graph Neural Networks

Intent-aware Graph Neural Network for Point-of-Interest Embedding and Recommendation

Graph-enhanced and collaborative attention networks for session-based recommendation

Enhanced graph neural network for session-based recommendation

Neural graph personalized ranking for Top-N Recommendation

Personalized Recommendation Via Inductive Spatiotemporal Graph Neural Network

A Time-Sensitive Graph Neural Network for Session-Based New Item Recommendation

Hybrid-Order Gated Graph Neural Network for Session-Based Recommendation

Sequence-Aware Graph Neural Network for Session-based Recommendation

Session-based Interactive Recommendation Via Deep Reinforcement Learning

Relevance-aware Graph Neural Network for Session-Based Recommendation

Hierarchical Transition-Aware Graph Attention Network for Session-based Recommendation

Session-based Social Recommendation via Dynamic Graph Attention Networks

Rethinking the Item Order in Session-based Recommendation with Graph Neural Networks