《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (6): 1869-1878.DOI: 10.11772/j.issn.1001-9081.2024060859

• 数据科学与技术 • 上一篇    

基于风味嵌入异构图层次学习的食谱推荐模型

颜文婧1,2, 王瑞东1,2, 左敏1,3(), 张青川1,2   

  1. 1.北京工商大学 农产品质量安全追溯技术及应用国家工程研究中心,北京 100048
    2.北京工商大学 计算机与人工智能学院,北京 100048
    3.北京物资学院 物流学院,北京 101149
  • 收稿日期:2024-06-25 修回日期:2024-09-05 接受日期:2024-09-06 发布日期:2024-10-08 出版日期:2025-06-10
  • 通讯作者: 左敏
  • 作者简介:颜文婧(1985—),女,安徽淮南人,讲师,博士,主要研究方向:生物信息智能处理、深度学习
    王瑞东(1999—),男,内蒙古赤峰人,硕士研究生,主要研究方向:生物信息智能处理、深度学习
    左敏(1973—),男,安徽铜陵人,教授,博士,主要研究方向:食品大数据、深度学习 zuomin@btbu.edu.cn
    张青川(1982—),男,河北石家庄人,副教授,博士,主要研究方向:数据挖掘、自然语言处理。
  • 基金资助:
    国家重点研发计划项目(2022YFF0606803);北京市属高校教师队伍建设支持计划高水平科研创新团队项目(BPHR20220104);北京学者项目(099);国家市场监督管理总局项目(2023MK169)

Recipe recommendation model based on hierarchical learning of flavor embedding heterogeneous graph

Wenjing YAN1,2, Ruidong WANG1,2, Min ZUO1,3(), Qingchuan ZHANG1,2   

  1. 1.National Engineering Research Centre for Agri-product Quality Traceability,Beijing Technology and Business University,Beijing 100048,China
    2.School of Computer and Artificial Intelligence,Beijing Technology and Business University,Beijing 100048,China
    3.Logistics School,Beijing Wuzi University,Beijing 101149,China
  • Received:2024-06-25 Revised:2024-09-05 Accepted:2024-09-06 Online:2024-10-08 Published:2025-06-10
  • Contact: Min ZUO
  • About author:YAN Wenjing, born in 1985, Ph. D., lecturer. Her research interests include intelligent processing of biological information, deep learning.
    WANG Ruidong, born in 1999, M. S. candidate. His research interests include intelligent processing of biological information, deep learning.
    ZUO Min, born in 1973, Ph. D., professor. His research interests include food big data, deep learning.
    ZHANG Qingchuan, born in 1982, Ph. D., associate professor. His research interests include data mining, natural language processing.
  • Supported by:
    National Key Research and Development Program of China(2022YFF0606803);High-level Research and Innovation Team Program of Beijing Municipal University Teacher Team Construction Support Plan(BPHR20220104);Beijing Scholars Program (099);State Administration for Market Regulation Project(2023MK169)

摘要:

针对食谱推荐任务中信息维度不全面、交互数据稀疏和交互信息冗余的问题,提出一种基于风味嵌入异构图层次学习的食谱推荐模型(RecipeFlavor)。首先,引入风味分子维度,并基于用户、食物、食材和食材的风味物质构建异构图,有效表示4种节点之间的联系;其次,基于信息传递机制构建基于异构图的层级学习模块,并结合压缩注意力(SA)机制,将节点的不同关系视为不同的信息通道,提取节点之间的关键交互信息并抑制噪声;最后,基于特征感知噪声构建对比学习(CL)模块,在模型学习中引入正负样本区分任务,增强用户和食谱节点之间的信息关联,提升模型对特征的学习能力。实验结果表明,RecipeFlavor在Recipe 1M+大型数据集上,与HGAT(Hierarchical Graph ATtention network for recipe recommendation)模型相比,曲线下面积(AUC)提升了1.44个百分点,Top-10的模型精确度(Pre)、命中率(HR)、平均精度(MAP)和归一化折损累计增益(NDCG)分别提升了0.76、6.11、2.68和3.05个百分点。可见,风味分子信息的引入拓展了食谱推荐的学习维度,而RecipeFlavor能有效提取异构图中的关键信息,增强用户和食谱之间的关联性,从而提升食谱推荐的精度。

关键词: 图卷积网络, 异构图学习, 压缩注意力机制, 推荐系统, 对比学习

Abstract:

Aiming at the problems of incomplete information dimension, sparse interaction data and redundant interaction information in recipe recommendation tasks, a Recipe recommendation model based on hierarchical learning of Flavor embedding heterogeneous graph (RecipeFlavor) was proposed. Firstly, the flavor molecule dimension was introduced, and a heterogeneous graph was constructed on the basis of users, foods, ingredients and flavor substances of ingredients to represent the connection among four kinds of nodes effectively. Then, a hierarchical learning module based on heterogeneous graph was constructed on the basis of information transmission mechanism, and combined with Squeeze Attention (SA) mechanism, different node relationships were regarded as different information channels, so that key interaction information between nodes was extracted and noise was suppressed. Finally, a Contrastive Learning (CL) module was constructed on the basis of feature-aware noise, and positive and negative sample discrimination tasks were introduced in model learning, thereby enhancing the information associations among users and recipe nodes and improving the model’s learning ability for features. Experimental results show that compared with HGAT (Hierarchical Graph ATtention network for recipe recommendation) model on Recipe 1M+ large dataset, RecipeFlavor has the Area Under the ROC Curve (AUC) increased by 1.44 percentage points, and the model Precision (Pre), Hit Rate (HR), Mean Average Precision (MAP), and Normalized Discounted Cumulative Gain (NDCG) of Top-10 increased by 0.76, 6.11, 2.68, and 3.05 percentage points, respectively. It can be seen that the introduction of flavor molecule information expands the learning dimension of recipe recommendation, and RecipeFlavor can extract key information in heterogeneous graph effectively, and enhance correlation among users and recipes, and thus improving the precision of recipe recommendations.

Key words: Graph Convolution Network (GCN), heterogeneous graph learning, Squeeze Attention (SA) mechanism, recommendation system, Contrastive Learning (CL)

中图分类号: