Recommendation Unlearning Algorithm Combining Fuzzy Clustering and Adaptive Denoising

doi:10.3778/j.issn.1673-9418.2312020

Abstract

Abstract: Privacy protection plays a crucial role in recommender systems as it helps to protect users’ sensitive information from disclosure risks. Recent recommendation unlearning has attracted increasing attention as an effective method of privacy protection. Existing methods often partition data into sub-partitions before training to enhance model training efficiency. However, simply partitioning interactions into sub-partitions can disrupt the integrity of user-item relationships and reduce the availability of data. In addition, the presence of false-positive noise in sub-partitions with implicit feedback can interfere with model training, preventing it from accurately capturing users’ true preferences. To address these challenges, a recommendation unlearning algorithm combining fuzzy clustering and adaptive denoising (FDRU) is proposed. Firstly, fuzzy clustering determines membership by calculating cosine distances between samples and various cluster centers, subsequently dividing the training dataset into several sub-partitions. Then, FDRU designs an adaptive denoising algorithm that dynamically eliminates false positive noise in sub-partitions based on thresholds. Finally, it utilizes dynamic weighted aggregation of sub-models for prediction and top-N recommendations. In order to assess the performance of the proposed algorithm, extensive experiments are carried out on three public datasets. Experimental results indicate that FDRU outperforms other benchmark algorithms on Recall and NDCG.

Key words: privacy protection, recommendation, unlearning, fuzzy clustering, adaptive denoising

摘要： 隐私保护在推荐系统中具有至关重要的地位，因为它有助于保护用户的敏感信息免受泄露风险。近年来，推荐遗忘学习作为一种有效的隐私保护手段引起了越来越多的关注。现有方法为了提高模型的训练效率，通常将数据划分为子分区进行训练。然而，简单划分子分区会破坏用户-项目间的完整性，降低数据的可用性。此外，子分区中隐式反馈的假阳性噪声会干扰模型的训练，使其无法准确地捕捉用户的真实偏好。为解决上述问题，提出了融合模糊聚类和自适应去噪的推荐遗忘学习算法（FDRU）。该算法使用模糊聚类来划分数据集，通过计算交互样本到各个聚类中心的余弦距离来确定隶属度，进而将训练集划分为若干个子分区。FDRU设计了一种自适应去噪方法，其能够根据阈值动态地剔除子分区中的假阳性噪声。通过动态权重聚合子模型进行预测和Top-N推荐。为了验证提出算法的性能，在三个公开数据集上进行实验验证，实验结果表明，提出的算法在召回率和归一化折损累计增益上优于其他基准算法。

关键词: 隐私保护, 推荐, 遗忘学习, 模糊聚类, 自适应去噪

WANG Jianfang, CHAI Guangwen, CHEN Yiqing, LIANG Menghao, LUO Junwei. Recommendation Unlearning Algorithm Combining Fuzzy Clustering and Adaptive Denoising[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(11): 2969-2979.

王建芳, 柴广文, 陈艺卿, 梁梦豪, 罗军伟. 融合模糊聚类和自适应去噪的推荐遗忘学习算法[J]. 计算机科学与探索, 2024, 18(11): 2969-2979.

References

[1] WAN Q, HE X N, WANG X, et al. Cross pairwise ranking for unbiased item recommendation[C]//Proceedings of the ACM Web Conference 2022, Apr 25-29, 2022. New York: ACM, 2022: 2370-2378.
[2] 冯晗, 伊华伟, 李晓会, 等. 推荐系统的隐私保护研究综述[J]. 计算机科学与探索, 2023, 17(8): 1814-1832.
FENG H, YI H W, LI X H, et al. Review of privacy-preserving research in recommendation systems[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(8): 1814-1832.
[3] GAO C, HUANG C, LIN D S, et al. DPLCF: differentially private local collaborative filtering[C]//Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, Jun 25-30, 2020. New York: ACM, 2020: 961-970.
[4] MINTO L, HALLER M, LIVSHITS B, et al. Stronger privacy for federated collaborative filtering with implicit feedback[C]//Proceedings of the 15th ACM Conference on Recommender Systems, Sep 1, 2021. New York: ACM, 2021: 342-350.
[5] CAO Y Z, F Y J. Towards making systems forget with machine unlearning[C]//Proceedings of the 2015 IEEE Symposium on Security and Privacy, May 17-21, 2015. Piscataway: IEEE, 2015: 463-480.
[6] BOURTOULE L, CHANDRASEKARAN V, CHOQUETTE-CHOO C A, et al. Machine unlearning[C]//Proceedings of the 2021 IEEE Symposium on Security and Privacy, May 24-27, 2021. Piscataway: IEEE, 2021: 141-159.
[7] CHEN M, ZHANG Z K, WANG T H, et al. Graph unlearning[C]//Proceedings of the 2022 ACM SIGSAC Conference on Computer and Communications Security, Nov 2022. New York: ACM, 2022: 499-513.
[8] CHEN C, SUN F, ZHANG M, et al. Recommendation unlearning[C]//Proceedings of the ACM Web Conference 2022, Apr 25-29, 2022. New York: ACM, 2022: 2768-2777.
[9] KAPOOR A, SINGHAL A. A comparative study of K-means, K-means++ and fuzzy C-means clustering algorithms[C]//Proceedings of the 2017 3rd International Conference on Computational Intelligence & Communication Technology, Mar 9-10, 2017. Piscataway: IEEE, 2017: 1-6.
[10] WEN H, YANG L, DEBORAH E. Leveraging post-click feedback for content recommendations[C]//Proceedings of the 13th ACM Conference on Recommender Systems, Sep 16, 2019. New York: ACM, 2019: 278-286.
[11] TIAN C X, XIE Y X, LI Y L, et al. Learning to denoise unreliable interactions for graph collaborative filtering[C]//Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, Jul 11-15, 2022. New York: ACM, 2022: 122-132.
[12] GAO Y J, DU Y T, HU Y J, et al. Self-guided learning to denoise for robust recommendation[C]//Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, Jul 11-15, 2022. New York: ACM, 2022: 1412-1422.
[13] CHEN H Y, LIN Y S, PAN M H, et al. Denoising self-attentive sequential recommendation[C]//Proceedings of the 16th ACM Conference on Recommender Systems, Sep 18-23, 2022. New York: ACM, 2022: 92-101.
[14] WANG W J, FENG F L, HE X N, et al. Denoising implicit feedback for recommendation[C]//Proceedings of the 14th ACM International Conference on Web Search and Data Mining, Mar 8-12, 2021. New York: ACM, 2021: 373-381.
[15] LU H Y, ZHANG M, MA W Z, et al. Effects of user negative experience in mobile news streaming[C]//Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, Jul 21-25, 2019. New York: ACM, 2019: 705-714.
[16] JAWAHEER G, SZOMSZOR M, KOSTKOVA P. Comparison of implicit and explicit feedback from an online music recommendation service[C]//Proceedings of the 1st International Workshop on Information Heterogeneity and Fusion in Recommender Systems. New York: ACM, 2010: 47-51.
[17] XU M, SUN J, YANG X, et al. Netflix and forget: efficient and exact machine unlearning from bi-linear recommendations[EB/OL]. [2023-10-23]. https://arxiv.org/abs/2302.06676.
[18] ZHANG C, CHEN R, ZHAO X Y, et al. Denoising and prompt-tuning for multi-behavior recommendation[C]//Proceedings of the ACM Web Conference 2023, Apr 25-29, 2023. New York: ACM, 2023: 1355-1363.
[19] LEE S. Fuzzy clustering with optimization for collaborative filtering-based recommender systems[J]. Journal of Ambient Intelligence and Humanized Computing, 2022, 13(9): 4189-4206.
[20] DUAN L, WANG W P, HAN B J. A hybrid recommendation system based on fuzzy C-means clustering and supervised learning[J]. KSII Transactions on Internet and Information Systems, 2021, 15(7): 2399-2413.
[21] HU Y J, KOREN Y, VOLINSKY C. Collaborative filtering for implicit feedback datasets[C]//Proceedings of the 2008 8th IEEE International Conference on Data Mining, Dec 15-19, 2008. Piscataway: IEEE, 2008: 263-272.
[22] RENDLE S, FREUDENTHALER C, GANTNER Z, et al. BPR: Bayesian personalized ranking from implicit feedback[C]//Proceedings of the 25th Conference on Uncertainty in Artificial Intelligence, Jun 18-21, 2009. Arlington: AUAI Press, 2009: 452-461.
[23] HE X N, DENG K, WANG X, et al. LightGCN: simplifying and powering graph convolution network for recommendation[C]//Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, Jun 25-30, 2020. New York: ACM, 2020: 639-648.
[24] MATUSZYK P, VINAGRE J, SPILIOPOULOU M, et al. Forgetting methods for incremental matrix factorization in recommender systems[C]//Proceedings of the 30th Annual ACM Symposium on Applied Computing, Apr 13-17, 2015. New York: ACM, 2023: 947-953.
[25] LI Y, CHEN C, ZHENG X, et al. Making recommender systems forget: learning and unlearning for erasable recommendation[J]. Knowledge-Based Systems, 2024, 283: 111124.
[26] LI Y, CHEN C, ZHANG Y, et al. UltraRE: enhancing RecEraser for recommendation Unlearning via error decomposition[C]//Advances in Neural Information Processing Systems 36, New Orleans, Dec 10-16, 2023.
[27] LIU B L, BAI B, XIE W, et al. Task-optimized user clustering based on mobile APP usage for cold-start recommendations[C]//Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Aug 14-18, 2022. New York: ACM, 2022: 3347-3356.
[28] FARAHANI M G, TORKESTANI J A, RAHMANI M. Adap-tive personalized recommender system using learning auto-mata and items clustering[J]. Information Systems, 2022, 105: 106101978.
[29] 王永贵, 李昕. 融合狼群算法和模糊聚类的混合推荐算法[J]. 计算机工程与应用, 2022, 58(5): 104-111.
WANG Y G, LI X. Hybrid recommendation algorithm combining wolf colony algorithm and fuzzy clustering[J]. Computer Engineering and Applications, 2022, 58(5): 104-111.
[30] LAVEE G, KOENIGSTEIN N, BARKAN O. When actions speak louder than clicks: a combined model of purchase probability and long-term customer satisfaction[C]//Procee-dings of the 13th ACM Conference on Recommender Systems, Sep 16, 2019. New York: ACM, 2019: 287-295.
[31] 张卫国, 袁炜轩, 周熙然. 融合深度去噪自编码器和注意力机制的推荐算法[J]. 计算机应用与软件, 2023, 40(8): 283-290.
ZHANG W G, YUAN W X, ZHOU X R. Recommendation algorithm combining deep denoising autoencoder and attention mechanism[J]. Computer Applications and Software, 2023, 40(8): 283-290.
[32] MCAULEY J, TARGETT C, SHI Q F, et al. Image-based recommendations on styles and substitutes[C]//Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval, Aug 9-13, 2015. New York: ACM, 2015: 43-52.