为了弥补基于用户身份识别和历史行为的传统推荐算法的不足，基于会话的推荐算法近年来受到了越来越多的关注。与传统的推荐算法不同，基于会话的推荐能够从用户最近的会话中捕获用户的短期偏好，从而获得更准确和及时的建议。但是目前基于会话的推荐算法在实际应用时往往遇到两个挑战：一个是在会话的表示学习过程中，噪声项目对特征提取有不可避免的影响，同时不同会话之间的内在关联需要进一步挖掘；另外一个主要挑战是在稀缺商品推荐或新建立的电商平台上的训练数据无法满足已有推荐模型的训练需求，不足以训练稳定鲁棒的推荐算法。因此，本文针对上述问题完成了以下工作：

     (1)为了降低噪声项目对会话表示学习的负面影响和挖掘不同会话之间的联系，本文提出了一种基于图双注意力模型的会话推荐算法。图双注意力模型的稀疏自注意力学习模块可以对用户错误点击项给定一个零值的注意力权重，以降低噪声项目对会话表示学习的负面影响。然后，图注意力学习模块结合双注意力会话表示和图卷积神经网络进一步挖掘不同会话样本之间的关系，得到更鲁棒的图双注意力会话表示。通过上述方式，可以为会话序列构建一个更可靠的会话表示，从而提高了推荐性能。

      (2)针对实际应用场景中训练样本不足的小样本会话推荐问题，本文提出了一种基于图注意力迁移学习模型的小样本会话推荐算法。图注意力迁移学习模型的会话内注意力表示学习模块重点挖掘会话内部关系，研究会话内每个项目对会话表示学习的重要性。然后，模型的跨域会话间表示学习模块对不同域之间的会话样本构建图关系，融合对抗迁移学习机制进行会话表示学习。模型学习到的会话表示融合了源域的会话数据特征，可以解决目标域的小样本会话推荐问题。

     (3)为了验证本文提出的两个算法的有效性，本文在Retailrocket和Diginetica两个公开数据集上进行了多组实验。实验结果表明：本文提出的图双注意力模型的HR@20和MRR@20指标均优于前沿基准模型；本文提出的图注意力迁移学习模型在解决基于会话推荐的小样本学习问题上有良好的性能。

    In order to make up for the deficiencies of traditional recommendation algorithms, which are based on user identity and historical behaviors, session-based recommendation algorithms have received more and more attention in recent years. Unlike traditional recommendation methods, session-based recommendation captures users' short-term preferences from their most recent sessions for more accurate and timely suggestions. However, the current session-based recommendation algorithm often encounters two challenges in its practical application: the one is that in the process of session representation learning, irrelevant items have an unavoidable impact on feature extraction, while the internal association between different sessions needs to be further explored; Another major challenge is that the training data on a scarce commodity recommendation or newly established commerce platform cannot meet the training needs of the existing recommendation models and cannot be sufficient to train robust and stable recommendation algorithms. Therefore, the following works have been done to solve the above problems:

    (1) In order to reduce the influence of noisy items on session representation learning and mine the relationship between different sessions, a session-based recommendation algorithm based on the graph dual attention model is proposed in this paper. The sparse self-attention learning module of the graph dual attention model can give a zero value attention weight to the user's wrong clicks, so as to reduce the negative impact of noise items on session representation learning. Then, the graph attention learning module combines the dual attention session representation and graph convolution neural network to further mine the relationship between different sessions, and obtain a more robust graph dual attention session representation. In this way, we can build a more reliable session representation for the session sequence, thus improving the recommendation performance.

    (2) Aiming at the problem of few-shot session recommendation with insufficient training samples in practical application scenarios, this paper proposes a few-shot session-based recommendation algorithm based on the graph attentive transfer learning model. The intra-session attention representation learning module in graph attentive transfer learning model focuses on mining the internal relationship between different items in each session and studying the importance of each item. The cross-domain inter-session representation learning module constructs the graph relationship for session samples between different domains. It carries out session representation learning in combination with the adversarial transfer learning mechanism. Thus, the learned session representation integrates the session data characteristics of the source domain and effectively solves the problem of few-shot session-based recommendation in the target domain.

    (3) To verify the validity of the two methods proposed in this thesis, a number of experiments have been carried out on two public datasets, Retailrocket and Diginetica. The experimental results show that the proposed graph dual attention model achieves better results in HR@20 and MRR@20 metrics than advanced benchmark models. The graph attentive transfer learning model proposed in this thesis performs well in solving few-shot learning session-based recommendation problem.

[1]于蒙,何文涛,周绪川,崔梦天,吴克奇,周文杰.推荐系统综述[J].计算机应用,2022, 42(4):1-16.

[2]张保淑.中国网民规模超10亿[N].人民日报海外版,2021-08-28(002).

[3]Gwadabe T R, Liu Y. IC-GAR: item co-occurrence graph augmented session-based recommendation[J]. Neural Computing and Applications, 2022, 34(10):7581-7596.

[4]Hui B, Zhang L, Zhou X, et al. Personalized recommendation system based on knowledge embedding and historical behavior[J]. Applied Intelligence, 2022, 52(1):954-966.

[5]刘华玲,马俊,张国祥. 基于深度学习的内容推荐算法研究综述[J].计算机工程, 2021, 47(07):1-12.

[6]王晨阳,任一,马为之,张敏,刘奕群,马少平.ReChorus:综合高效易扩展的轻量级推荐算法框架[J].软件学报, 2022,33(04):1430-1438.

[7]刘君良,李晓光.个性化推荐系统技术进展[J].计算机科学,2020,47(07):47-55.

[8]Tahmasebi F, Meghdadi M, Ahmadian S, et al. A hybrid recommendation system based on profile expansion technique to alleviate cold start problem[J]. Multimedia Tools and Applications, 2021, 80(2): 2339-2354.

[9]Sharma S, Rana V, Malhotra M. Automatic recommendation system based on hybrid filtering algorithm[J]. Education and Information Technologies, 2022, 27(2): 1523-1538.

[10]Dhelim S, Aung N, Bouras M A, et al. A survey on personality-aware recommendation systems[J]. Artificial Intelligence Review, 2022, 55(3): 2409-2454.

[11]赵海燕,赵佳斌,陈庆奎,等.会话推荐系统[J].小型微型计算机系统, 2019,40(9):1869-1875.

[12]陈聪,张伟,王骏. 带有时间预测辅助任务的会话式序列推荐[J]. 计算机学报, 2021, 44(9):1841-1853.

[13]Xia X, Yin H, Yu J, et al. Self-supervised hypergraph convolutional networks for session-based recommendation[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2021, 35(5): 4503-4511.

[14]Ludewig M, Mauro N, Latifi S, et al. Empirical analysis of session-based recommendation algorithms[J]. User Modeling and User-Adapted Interaction, 2021, 31(1): 149-181.

[15]Zang T, Zhu Y, Zhu J, et al. MPAN: Multi-parallel attention network for session-based recommendation[J]. Neurocomputing, 2022, 471: 230-241.

[16]Sarwar B, Karypis G, Konstan J, et al. Item-based collaborative filtering recommendation al-gorithms [C]. // Proceedings of the 10th international conference on World Wide Web, 2001:285–295.

[17]Linden G, Smith B, York J. Amazon. com recommendations: Item-to-item collaborative filter-ing [J]. IEEE Internet Computing, 2003, 7 (1): 76–80.

[18]Shani G, Heckerman D, Brafman R I, et al. An MDP-based recommender system. [J]. Journal ofMachine Learning Research, 2005, 6 (9): 1265–1295.

[19]Rendle S, Freudenthaler C, Schmidt-Thieme L. Factorizing personalized markov chains for next-basket recommendation[C]//Proceedings of the 19th international conference on World wide web. 2010: 811-820.

[20]Yang B, Lei Y, Liu D Y. et al. Social collaborative filtering by trust [C]//Proceedings of the 23rd International Joint Conference on Artificial Intelligence. Beijing, China, 2013: 2747–2753.

[21]He X, Zhang H, Kan M Y, et al. Fast matrix factorization for online recommendation with implicit feedback[C]//Proceedings of the 39th International ACM SIGIR conference on Research and Development in Information Retrieval. ACM, 2016: 549-558.

[22]Hidasi B, Karatzoglou A. Recurrent neural networks with top-k gains for session-based recom-mendations [C]. // Proceedings of the 27th ACM international conference on information andknowledge management, 2018: 843–852.

[23]Quadrana M, Karatzoglou A, Hidasi B, et al. Personalizing session-based recommendations with hierarchical recurrent neural networks[C]//proceedings of the Eleventh ACM Conference on Recommender Systems. 2017: 130-137.

[24]Gwadabe T R, Liu Y. Improving graph neural network for session-based recommendation system via non-sequential interactions[J]. Neurocomputing, 2022, 468: 111-122.

[25]Wu S, Tang Y, Zhu Y, et al. Session-based recommendation with graph neural networks [C]. //Proceedings of the AAAI Conference on Artificial Intelligence, 2019: 346–353.

[26]Xu C, Zhao P, Liu Y, et al. Graph Contextualized Self-Attention Network for Session-basedRecommendation. [C]. // IJCAI, 2019: 3940–3946.

[27]Qiu R, Huang Z, Li J, et al. Exploiting cross-session information for session-based recommendation with graph neural networks[J]. ACM Transactions on Information Systems (TOIS), 2020, 38(3): 1-23.

[28]Wang Z, Wei W, Cong G, et al. Global context enhanced graph neural networks for session-based recommendation[C]//Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. 2020: 169-178.

[29]Liu Q, Zeng Y, Mokhosi R, et al. STAMP: short-term attention/memory priority model for session-based recommendation[C]//Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2018: 1831-1839.

[30]Xu C, Zhao P, Liu Y, et al. Graph Contextualized Self-Attention Network for Session-based Recommendation[C]//IJCAI. 2019, 19: 3940-3946.

[31]Luo A, Zhao P, Liu Y, et al. Collaborative Self-Attention Network for Session-based Recommendation[C]//IJCAI. 2020: 2591-2597.

[32]Yu F, Zhu Y, Liu Q, et al. TAGNN: Target attentive graph neural networks for session-based recommendation[C]//Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. 2020: 1921-1924.

[33]Wang S, Cao L, Wang Y, et al. A survey on session-based recommender systems[J]. ACM Computing Surveys (CSUR), 2021, 54(7): 1-38.

[34]Berrone S, Santa F D, Mastropietro A, et al. Graph-Informed Neural Networks for Regressions on Graph-Structured Data[J]. Mathematics, 2022, 10(5): 1-29.

[35]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(01):47-80.

[36]Wu Z, Pan S, Chen F, et al. A comprehensive survey on graph neural networks[J]. IEEE transactions on neural networks and learning systems, 2020, 32(1): 4-24.

[37]Wu S, Sun F, Zhang W, et al. Graph neural networks in recommender systems: a survey[J]. arXiv preprint arXiv:2011.02260, 2020.

[38]任欢, 王旭光. 注意力机制综述[J]. 计算机应用, 2021, 41(S01): 1-6.

[39]Niu Z, Zhong G, Yu H. A review on the attention mechanism of deep learning[J]. Neurocomputing, 2021, 452: 48-62.

[40]Chaudhari S, Mithal V, Polatkan G, et al. An attentive survey of attention models[J]. ACM Transactions on Intelligent Systems and Technology (TIST), 2021, 12(5): 1-32.

[41]黄立威,江碧涛,吕守业,刘艳博,李德毅.基于深度学习的推荐系统研究综述[J].计算机学报,2018,41(07):1619-1647.

[42]Martins A, Astudillo R. From softmax to sparsemax: A sparse model of attention and multi-label classification[C]//International conference on machine learning. PMLR, 2016: 1614-1623.

[43]Correia G M, Niculae V, Martins A F T. Adaptively sparse transformers[J]. arXiv preprint arXiv:1909.00015, 2019.

[44]Tsallis C. Possible generalization of Boltzmann-Gibbs statistics[J]. Journal of statistical physics, 1988, 52(1): 479-487.

[45]Sun F, Liu J, Wu J, et al. BERT4Rec: Sequential recommendation with bidirectional encoder representations from transformer[C]//Proceedings of the 28th ACM international conference on information and knowledge management. 2019: 1441-1450.

[46]Jiang B, Zhang Z,Lin D,et al. Semi-supervised learning with graph learning-convolutional networks[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 11313-11320.

[47]Klambauer G, Unterthiner T, Mayr A, et al. Self-normalizing neural networks [C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. 2017: 972-981.

[48]Jannach D, Ludewig M. When recurrent neural networks meet the neighborhood for session-based recommendation[C]//Proceedings of the eleventh ACM conference on recommender systems. 2017: 306-310.

[49]Garg D, Gupta P, Malhotra P, et al. Sequence and time aware neighborhood for session-based recommendations: Stan[C]//Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. 2019: 1069-1072.

[50]Li J, Ren P, Chen Z,et al. Neural attentive session-based recommendation[C]//Proceedings of the 2017 ACM on Conference on Information and Knowledge Management. 2017: 1419-1428.

[51]Yuan J, Song Z, Sun M, et al. Dual Sparse Attention Network For Session-based Recommendation[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2021, 35(5): 4635-4643.

[52]Zhang Z, Wang B. Fusion of latent categorical prediction and sequential prediction for session-based recommendation[J]. Information Sciences, 2021, 569: 125-137.

[53]Xia X, Yin H, Yu J, et al. Self-supervised hypergraph convolutional networks for session-based recommendation[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2021, 35(5): 4503-4511.

[54]Zhang Z, Wang B. Graph Spring Network and Informative Anchor Selection for Session-based Recommendation[J]. arXiv preprint arXiv:2202.09502, 2022.

[55]赵凯琳, 靳小龙, 王元卓. 小样本学习研究综述[J]. 软件学报, 2021, 32(2): 349-369.

[56]Wang Y, Yao Q, Kwok J T, et al. Generalizing from a few examples: A survey on few-shot learning[J]. ACM computing surveys (csur), 2020, 53(3): 1-34.

[57]Yang X, Nan X, Song B. D2N4: A discriminative deep nearest neighbor neural network for few-shot space target recognition[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(5): 3667-3676.

[58]Tseng H Y, Lee H Y, Huang J B, et al. Cross-domain few-shot classification via learned feature-wise transformation[J]. arXiv preprint arXiv:2001.08735, 2020.

[59]Zhang B, Titov I, Sennrich R. Sparse Attention with Linear Units[J]. arXiv preprint arXiv:2104.07012, 2021.

[60]Sun H, Xu J, Zheng K, et al. MFNP: A Meta-optimized Model for Few-shot Next POI Recommendation[C]//IJCAI. 2021: 3017-3023.

[61]Sreepada R S, Patra B K. Mitigating long tail effect in recommendations using few shot learning technique[J]. Expert Systems with Applications, 2020, 140: 112887.

[62]Zhuang F, Qi Z, Duan K, et al. A comprehensive survey on transfer learning[J]. Proceedings of the IEEE, 2020, 109(1): 43-76.

[63]王坤峰, 苟超, 段艳杰, 等. 生成式对抗网络 GAN 的研究进展与展望[J]. 自动化学报, 2017, 43(3): 321-332.

[64]Goodfellow I, Pouget-Abadie J, Mirza M, et al. Generative Adversarial Networks[J]. COMMUNICATIONS OF THE ACM, 2020, 63(11): 1-9.

[65]Jin X, Park Y, Maddix D C, et al. Domain Adaptation for Time Series Forecasting via Attention Sharing[J]. arXiv preprint arXiv:2102.06828, 2021.

[66]Zhao J, Li H, Qu L, et al. DCFGAN: An adversarial deep reinforcement learning framework with improved negative sampling for session-based recommender systems[J]. Information Sciences, 2022, 596: 222-235.