Researchers from the Center for Research on Intelligent Perception and Computing, Institute of Automation of the Chinese Academy of Sciences (CASIA), and their collaborators proposed and publicized a session-based recommendation system with graph neural networks, SR-GNN in brevity. Related research was published at the world top academic conference of Artificial Intelligence - AAAI 2019.

Session is a mechanism used to recognize and record user profile. A session-based recommendation system can help users select certain items based on their constant behaviors in Web applications within a period of time. For instance, when someone consecutively clicks 10 goods at an online shopping application, then the 10 goods can constitute a short session. Due to the emergence of massive and anonymous information on the Internet, session-based recommendation is important and widely discussed.

Previous methods model a session as a sequence and estimate user representations besides item representations to make recommendations. They are insufficient to obtain accurate user vectors in sessions and neglect complex transitions of items.

To obtain accurate item embedding and take complex transitions of items into account, session sequences are modeled as graph structured data in the proposed method - SR-GNN for short. Based on the session graph, GNN can capture complex transitions of items, which are difficult to be revealed by previous conventional sequential methods.

Each session is then represented as the composition of the global preference and the current interest of that session using an attention network. Extensive experiments conducted on two real datasets show that SR-GNN evidently outperforms the state-of-the-art session-based recommendation methods consistently.

The Workflow of the SR-GNN Method (Image by CASIA)