Journal of Computer Applications ›› 2021, Vol. 41 ›› Issue (2): 330-336.DOI: 10.11772/j.issn.1001-9081.2020060805

Special Issue: 人工智能

• Artificial intelligence • Previous Articles     Next Articles

Multi-graph neural network-based session perception recommendation model

NAN Ning1,2, YANG Chengyi2,3, WU Zhihao1,2   

  1. 1. School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044, China;
    2. Key Laboratory of Intelligent Passenger Service of Civil Aviation, Civil Aviation Administration of China, Beijing 100105, China;
    3. China Civil Aviation Information Network Incorporation Limited, Beijing 101318, China
  • Received:2020-06-15 Revised:2020-09-29 Online:2021-02-10 Published:2020-12-18
  • Supported by:
    This work is partially supported by the Science and Technology Major Project of Civil Aviation Administration of China (MHRD20160109).


南宁1,2, 杨程屹2,3, 武志昊1,2   

  1. 1. 北京交通大学 计算机与信息技术学院, 北京 100044;
    2. 中国民用航空局 民航旅客服务智能化应用技术重点实验室, 北京 100105;
    3. 中国民航信息网络股份有限公司, 北京 101318
  • 通讯作者: 杨程屹
  • 作者简介:南宁(1995-),男,山东泰安人,硕士研究生,主要研究方向:机器学习、数据挖掘;杨程屹(1986-),河北唐山人,工程师,博士,主要研究方向:人工智能、收益管理、民航智能化;武志昊(1984-),山西大同人,副教授,博士生导师,博士,主要研究方向:数据与知识工程、人工智能。
  • 基金资助:

Abstract: The session-based recommendation algorithms mainly rely on the information from the target session, but fail to fully utilize the collaborative information from other sessions. In order to solve this problem, a Multi-Graph neural network-based Session Perception recommendation (MGSP) model was proposed. Firstly, according to the target session and all sessions in the training set, Item-Transition Graph (ITG) and Collaborative Relation Graph (CRG) were constructed. Based on these two graphs, the Graph Neural Network (GNN) was applied to aggregate the information of the nodes in order to obtain two types of node representations. Then, after the two-layer attention module modelling two type node representations, the session-level representation was obtained. Finally, by using the attention mechanism to fuse the information, the ultimate session representation was gained, and the next interaction item was predicted. The comparison experiments were carried out in two scenarios of e-commerce and civil aviation. Experimental results show that, the proposed algorithm is superior to the optimal benchmark model, with an increase of more than 1 percentage point and 3 percentage point in the indicators on the e-commerce and civil aviation datasets respectively, verifying the effectiveness of the proposed model.

Key words: session-based recommendation, multi-graph neural network, attention mechanism, personal preference, collaborative information

摘要: 针对基于会话的推荐算法主要依赖目标会话中的信息,而未充分利用其他会话中的协同信息的问题,提出了一种基于多图神经网络的会话感知推荐(MGSP)模型。首先,根据目标会话与训练集中的所有会话构建物品转移图(ITG)和协同关联图(CRG),并基于这两张图应用图神经网络(GNN)来汇聚节点的信息,得到两类的节点表示;然后,经过双层注意力模块对两类节点表示建模,获取会话级别的表示;最后,使用注意力机制进行信息融合,得到最终的会话表示,并预测下一个交互物品。分别在电商和民航两个场景下进行了对比实验,实验结果表明,相较最优的基准模型,MGSP模型在电商数据集各项指标上的提高超过1个百分点,在民航数据集各项指标上的提高约为3个百分点,验证了MGSP模型的有效性。

关键词: 基于会话的推荐, 多图神经网络, 注意力机制, 个性化偏好, 协同信息

CLC Number: