基于图神经网络的异构信任推荐算法

doi:10.19678/j.issn.1000-3428.0062235

摘要/Abstract

摘要： 传统基于图神经网络的社交推荐算法通过加强用户和项目特征的学习提升预测精度，但随着用户数据日益稀疏和社交关系趋于复杂，推荐质量提升缓慢。为挖掘用户和项目的潜在关联关系，提出一种结合图神经网络的异构信任推荐算法（GraphTrust）。在显式信任关系的基础上获取用户的潜在好友，根据动态影响力传播模型将图神经网络中的节点和边进行分类，通过不同类型的边在不同节点间进行影响力传播扩散，捕捉隐藏在高阶网络结构中的影响力扩散特征，并使用户和项目的潜在特征随着影响力传播过程达到平衡状态，最终将用户交互的项目特征作为辅助特征与用户特征聚合进行评分预测。在Yelp和Flickr数据集上的实验结果表明，当潜在特征维数为64时，GraphTrust算法相比于DiffNet++算法的命中率和归一化折损累计增益分别提升了13.2%、22.2%和20.4%、25.5%，在一定程度上提高了推荐过程的可解释性和预测精度，并且缓解了数据稀疏问题。

Abstract: Traditional social recommendation algorithms based on graph neural networks improve prediction accuracy by enhancing the learning of user and item features;however, the recommendation quality slowly decreases as the data about users become increasingly sparse, and social relationships tend to be complex.To explore the potential relationship between users and items, this study proposes a heterogeneous trust recommendation algorithm combining graph neural networks(GraphTrust).The algorithm obtains potential friends of users on the basis of explicit trust relationships. According to the dynamic influence propagation model, the nodes and edges in a graph neural network are classified, and influence propagation and diffusion are carried out between different nodes through different types of edges, to capture the influence diffusion features hidden in the high-order network structure.Additionally, this is to enable the potential characteristics of users and items to reach a balance with the influence dissemination process.Finally, the item features interacted with by users are aggregated as auxiliary and user features for scoring prediction.The experimental results show that when the potential feature dimension is 64, the Hit Rate(HR) of GraphTrust on Yelp and Flickr datasets is increased by 13.2%, 22.2% compared with a neural influence and interest diffusion network for social recommendation(DiffNet++), and the Normalized Discounted Cumulative Gain(NDCG) of GraphTrust is increased by 20.4%, 25.5%, which improves the interpretability and prediction accuracy of the recommendation process to a certain extent, and alleviates the problem of data sparsity.

Key words: recommendation algorithm, social trust relationship, influence spread, feature representation, potential feature

中图分类号:

TP311

徐上上, 孙福振, 王绍卿, 董家玮, 吴田慧. 基于图神经网络的异构信任推荐算法[J]. 计算机工程, 2022, 48(9): 89-95,104.

XU Shangshang, SUN Fuzhen, WANG Shaoqing, DONG Jiawei, WU Tianhui. Heterogeneous Trust Recommendation Algorithm Based on Graph Neural Networks[J]. Computer Engineering, 2022, 48(9): 89-95,104.

https://www.ecice06.com/CN/Y2022/V48/I9/89

图/表 9

20220924175353

20220924175357

20220924175401

20220924175405

20220924175408

20220924175411

20220924175415

20220924175418

20220924175422

参考文献

[1] SALAKHUTDINOV R, MNIH A.Probabilistic matrix factorization[C]//Proceedings of International Conference on Neural Information Processing Systems.[S.l.]:Curran Associates Inc., 2007:1257-1264.
[2] AHARON M, ELAD M, BRUCKSTEIN A.K-SVD:an algorithm for designing overcomplete dictionaries for sparse representation[J].IEEE Transactions on Signal Processing, 2006, 54(11):4311-4322.
[3] KOREN Y.Factorization meets the neighborhood:a multifaceted collaborative filtering model[C]//Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.New York, USA:ACM Press, 2008:426-434.
[4] JAMALI M, ESTER M.A matrix factorization technique with trust propagation for recommendation in social networks[C]//Proceedings of the 4th ACM Conference on Recommender Systems.New York, USA:ACM Press, 2010:135-142.
[5] GUO G B, ZHANG J, YORKE-SMITH N.TrustSVD:collaborative filtering with both the explicit and implicit influence of user trust and of item ratings[C]//Proceedings of the 29th AAAI Conference on Artificial Intelligence. Palo Alto, USA:AAAI Press, 2015:123-129.
[6] 孟祥福, 张霄雁, 唐延欢, 等.基于地理-社会关系的多样性与个性化兴趣点推荐[J].计算机学报, 2019, 42(11):2574-2590. MENG X F, ZHANG X Y, TANG Y H, et al.A diversified and personalized recommendation approach based on geo-social relationships[J].Chinese Journal of Computers, 2019, 42(11):2574-2590.(in Chinese)
[7] 王刚, 蒋军, 王含茹, 等.基于联合概率矩阵分解的群推荐方法研究[J].计算机学报, 2019, 42(1):98-110. WANG G, JIANG J, WANG H R, et al.Study of group recommendation based on probabilistic matrix factorization[J].Chinese Journal of Computers, 2019, 42(1):98-110.(in Chinese)
[8] 刘峰, 王宝亮, 邹荣宇, 等.基于随机游走的网络表示学习推荐算法[J].计算机工程, 2021, 47(9):90-96, 105. LIU F, WANG B L, ZOU R Y, et al.Recommendation algorithm using network representation learning based on random walk[J].Computer Engineering, 2021, 47(9):90-96, 105.(in Chinese)
[9] YADAV A, CHAKRAVERTY S, SIBAL R.A survey of implicit trust on social networks[C]//Proceedings of International Conference on Green Computing and Internet of Things(ICGCIoT).Washington D.C., USA:IEEE Press, 2016:1511-1515.
[10] OARD D, KIM J.Implicit feedback for recommender systems[C]//Proceedings of AAAI Workshop on Recommender Systems.Palo Alto, USA:AAAI Press, 1998:81-83.
[11] ZHANG C X, YU L, WANG Y, et al.Collaborative user network embedding for social recommender systems[C]//Proceedings of 2017 SIAM International Conference on Data Mining.Philadelphia, USA:Society for Industrial and Applied Mathematics, 2017:381-389.
[12] LING Z H, XIAO Y Y, WANG H Y, et al.Extracting implicit friends from heterogeneous information network for social recommendation[M].Berlin, Germany:Springer, 2019.
[13] AHMADIAN S, JOORABLOO N, JALILI M, et al.A social recommender system based on reliable implicit relationships[J].Knowledge-Based Systems, 2020, 192:105371.
[14] 马帅, 刘建伟, 左信.图神经网络综述[J].计算机研究与发展, 2022, 59(1):47-80. MA S, LIU J W, ZUO X.Survey on graph neural network[J].Journal of Computer Research and Development, 2022, 59(1):47-80.(in Chinese)
[15] 刘志鑫, 张泽华, 张杰.基于多层次多视角的图注意力Top-N推荐方法[J].计算机科学, 2021, 48(4):104-110. LIU Z X, ZHANG Z H, ZHANG J.Top-N recommendation method for graph attention based on multi-level and multi-view[J].Computer Science, 2021, 48(4):104-110.(in Chinese)
[16] 何昊晨, 张丹红.基于多维社交关系嵌入的深层图神经网络推荐方法[J].计算机应用, 2020, 40(10):2795-2803. HE H C, ZHANG D H.Recommendation method based on multidimensional social relationship embedded deep graph neural network[J].Journal of Computer Applications, 2020, 40(10):2795-2803.(in Chinese)
[17] WANG X, HE X N, WANG M, et al.Neural graph collaborative filtering[C]//Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval.New York, USA:ACM Press, 2019:165-174.
[18] FAN W Q, MA Y, LI Q, et al.Graph neural networks for social recommendation[C]//Proceedings of 2019 World Wide Web Conference.New York, USA:ACM Press, 2019:417-426.
[19] WU L, SUN P J, FU Y J, et al.A neural influence diffusion model for social recommendation[C]//Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval.New York, USA:ACM Press, 2019:235-244.
[20] 张浩博, 薛峰, 刘凯.基于半自动编码器的协同过滤推荐算法[J].计算机工程, 2021, 47(3):125-130. ZHANG H B, XUE F, LIU K.Collaborative filtering recommendation algorithm based on semi-autoencoder[J].Computer Engineering, 2021, 47(3):125-130.(in Chinese)
[21] 李琳, 唐守廉.基于多层注意力表示的音乐推荐模型[J].电子学报, 2020, 48(9):1672-1679. LI L, TANG S L.Hierarchical attention representation model for music recommendation[J].Acta Electronica Sinica, 2020, 48(9):1672-1679.(in Chinese)
[22] WU Q T, ZHANG H R, GAO X F, et al.Dual graph attention networks for deep latent representation of multifaceted social effects in recommender systems[C]//Proceedings of 2019 World Wide Web Conference.New York, USA:ACM Press, 2019:2091-2102.
[23] WU L, LI J W, SUN P J, et al.DiffNet++:a neural influence and interest diffusion network for social recommendation[EB/OL].[2021-07-11].https://arxiv.org/abs/2002.00844.
[24] SHI C, HU B B, ZHAO W X, et al.Heterogeneous information network embedding for recommendation[J].IEEE Transactions on Knowledge and Data Engineering, 2019, 31(2):357-370.

选择文件类型/文献管理软件名称

选择包含的内容