将行为依赖融入多任务学习的个性化推荐模型

doi:10.3778/j.issn.1673-9418.2208098

摘要/Abstract

摘要： 多种类型行为数据的引入缓解了协同过滤算法存在的数据稀疏和冷启动问题，在推荐领域被广泛研究和应用。尽管当前对多行为推荐的研究已经取得很大进展，但仍然存在以下问题：未能全面捕获行为之间复杂的依赖关系；忽略了行为特征与用户和项目的相关性。这导致学习到的特征向量无法准确表达用户的兴趣偏好，使得推荐结果存在偏差。为了解决以上问题，提出了将行为依赖融入多任务学习的个性化推荐模型（BDMR），将行为之间复杂的依赖关系分为特征相关性和时序相关性。首先，设置用户个性化行为向量，利用图神经网络处理多个单行为交互图，联合用户、项目和行为特征聚合高阶邻域信息，结合注意力机制学习行为之间的特征相关性；其次，将行为特征和项目特征构成的交互序列输入长短期记忆网络，捕获行为之间的时序相关性；最后，将个性化行为向量融入多任务学习框架获取更加准确的用户、行为和项目特征。为了验证提出模型的性能，在三个真实数据集上进行对比实验，在Yelp数据集上，相较于最优基线，HR和NDCG分别提升了1.5%和2.9%；在ML20M数据集上，HR和NDCG分别提升了2.0%和0.5%；在Tmall数据集上，HR和NDCG分别提升了25.6%和30.2%。实验结果表明，该模型优于其他的基准模型。

关键词: 多行为推荐, 图神经网络, 循环神经网络, 多任务学习框架

Abstract: The introduction of multiple types of behavioral data alleviates the data sparsity and cold-start problems of collaborative filtering algorithms, which is widely studied and applied in the field of recommendations. Although great progress has been made in the current research on multi-behavior recommendation, the following problems still exist: failure to comprehensively capture the complex dependencies between behaviors; ignoring the relevance of behavior features to users and items, and the recommendation results are biased. This results in the learned feature vectors failing to accurately represent the user??s interest preferences. To solve the above problems, a person-alized recommendation model (BDMR) that integrates behavioral dependencies into multi-task learning is proposed, and in this paper, the complex dependencies between behaviors are divided into feature relevance and temporal relevance. Firstly, the user personalized behavior vector is set, and multiple interaction graphs are processed with graph neural networks which combine user, item and behavior features to aggregate higher-order neighborhood information, and attention mechanism is combined to learn feature relevance among behaviors. Secondly, the interaction sequence composed of behavior features and item features is input into a long and short-term memory network to capture the temporal relevance among behaviors. Finally, personalized behavior vectors are integrated into a multi-task learning framework to obtain more accurate user, behavior and item features. To verify the perf-ormance of this model, experiments are conducted on three real datasets. On the Yelp dataset, compared with the optimal baseline, HR and NDCG are improved by 1.5% and 2.9% respectively. On the ML20M dataset, HR and NDCG are increased by 2.0% and 0.5% respectively. On the Tmall dataset, HR and NDCG are improved by 25.6% and 30.2% respectively. Experimental results show that the model proposed in this paper is superior to baselines.

Key words: multi-behavior recommendation, graph neural networks, recurrent neural networks, multi-task learning framework

顾军华, 李宁宁, 王鑫鑫, 张素琪. 将行为依赖融入多任务学习的个性化推荐模型[J]. 计算机科学与探索, 2024, 18(1): 231-243.

GU Junhua, LI Ningning, WANG Xinxin, ZHANG Suqi. Integrating Behavioral Dependencies into Multi-task Learning for Personalized Recommendations[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(1): 231-243.

参考文献

[1] HE X N, LIAO L Z, ZHANG H W, et al. Neural colla-borative filtering[C]//Proceedings of the 26th International Conference on World Wide Web, Perth, Apr 3-7, 2017. New York: ACM, 2017: 173-182.
[2] GAO C, HE X N, GAN D H, et al. Learning to recommend with multiple cascading behaviors[J]. IEEE Transactions on Knowledge and Data Engineering, 2019, 33(6): 2588-2601.
[3] WANG W, ZHANG W, LIU S K, et al. Beyond clicks: mod-eling multi-relational item graph for session-based target behavior prediction[C]//Proceedings of the Web Conference 2020, Taipei, China, Apr 20-24, 2020. New York: ACM, 2020: 3056-3062.
[4] TANJIM M M, SU C, BENJAMIN E, et al. Attentive sequ-ential models of latent intent for next item recommendation[C]//Proceedings of the Web Conference 2020, Taipei, China，Apr 20-24, 2020. New York: ACM, 2020: 2528-2534.
[5] JIN B, GAO C, HE X, et al. Multi-behavior recommen-dation with graph convolutional networks[C]//Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, Jul 25-30, 2020. New York: ACM, 2020: 659-668.
[6] CHEN C, MA W, ZHANG M, et al. Graph heterogeneous multi-relational recommendation[C]//Proceedings of the 35th AAAI Conference on Artificial Intelligence, the 33rd Con-ference on Innovative Applications of Artificial Intell-igence, the 11th Symposium on Educational Advances in Artificial Intelligence, Feb 2-9, 2021. Menlo Park: AAAI, 2021: 3958-3966.
[7] NI Y, OU D, LIU S, et al. Perceive your users in depth: learning universal user representations from multiple e-commerce tasks[C]//Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, London, Aug 19-23, 2018. New York: ACM, 2018: 596-605.
[8] GUO L, HUA L, JIA R, et al. Buying or browsing? Pred-icting real-time purchasing intent using attention-based deep network with multiple behavior[C]//Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Anchorage, Aug 4-8, 2019. New York: ACM, 2019: 1984-1992.
[9] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.
[10] TANG Y, HUANG Y, WU Z, et al. Question detection from acoustic features using recurrent neural network with gated recurrent unit[C]//Proceedings of the 2016 IEEE Intern-ational Conference on Acoustics, Speech and Signal Proc-essing, Shanghai, Mar 20-25, 2016. Piscataway: IEEE, 2016: 6125-6129.
[11] XIA L, HUANG C, XU Y, et al. Knowledge-enhanced hierarchical graph transformer network for multi-behavior recommendation[C]//Proceedings of the 35th AAAI Confe-rence on Artificial Intelligence, the 33rd Conference on Innovative Applications of Artificial Intelligence, the 11th Symposium on Educational Advances in Artificial Intell-igence, Feb 2-9, 2021. Menlo Park: AAAI, 2021: 4486-4493.
[12] 郭晓旺, 夏鸿斌, 刘渊. 融合知识图谱与图卷积网络的混合推荐模型[J]. 计算机科学与探索, 2022, 16(6): 1343-1353.
GUO X W, XIA H B, LIU Y. Hybrid recommendation model of knowledge graph and graph convolutional net-work[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(6): 1343-1353.
[13] KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[C]//Proceedings of the 5th International Conference on Learning Representations, Toulon, Apr 24-26, 2017: 1-14.
[14] HAMILTON W, YING Z T, LESKOVEC J. Inductive representation learning on large graphs[C]//Advances in Neural Information Processing Systems 30, Long Beach, Dec 4-9,?2017: 1024-1034.
[15] WANG X, JI H, SHI C, et al. Heterogeneous graph atte-ntion network[C]//Proceedings of the 2019 World Wide Web Conference, San Francisco, May 13-17, 2019. New York: ACM, 2019: 2022-2032.
[16] DONKERS T, LOEPP B, ZIEGLER J. Sequential user-based recurrent neural network recommendations[C]//Proce-edings of the 11th ACM Conference on Recommender Systems, Como, Aug 27-31, 2017. New York: ACM, 2017: 152-160.
[17] LIU J, WU C, WANG J. Gated recurrent units based neural network for time heterogeneous feedback recommendation[J]. Information Sciences, 2018, 423: 50-65.
[18] ZHANG Y, YANG Q. An overview of multi-task learning[J]. National Science Review, 2018, 5(1): 30-43.
[19] CHEN Z, WANG X, XIE X, et al. Co-attentive multi-task learning for explainable recommendation[C]//Proceedings of the 28th International Joint Conference on Artificial Intelligence, Macao, China, 2019: 2137-2143.
[20] CHEN C, ZHANG M, ZHANG Y, et al. Efficient hetero-geneous collaborative filtering without negative sampling for recommendation[C]//Proceedings of the 34th AAAI Conference on Artificial Intelligence, the 32nd Innovative Applications of Artificial Intelligence Conference, the 10th AAAI Symposium on Educational Advances in Artificial Intelligence, New York, Feb 7-12, 2020. Menlo Park: AAAI, 2020: 19-26.
[21] XIA L, XU Y, HUANG C, et al. Graph meta network for multi-behavior recommendation[C]//Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, Jul 11-15, 2021. New York: ACM, 2021: 757-766.