Please wait a minute...
浙江大学学报(工学版)
Journal of ZheJiang University (Engineering Science)  2024, Vol. 58 Issue (8): 1717-1727    DOI: 10.3785/j.issn.1008-973X.2024.08.018
    
Cross-domain recommendation model based on source domain data augmentation and multi-interest refinement transfer
Yabo YIN(),Xiaofei ZHU*(),Yidan LIU
College of Computer Science and Engineering, Chongqing University of Technology, Chongqing 400054, China
Download: HTML     PDF(2143KB) HTML
Export: BibTeX | EndNote (RIS)      

Abstract  

A cross-domain recommendation model that utilizes source domain data augmentation and multi-interest refinement transfer was proposed in order to address the issues of difficulty in modeling interest preferences in cross-domain recommendation tasks caused by the lack of user interaction data in the source domain, as well as the problem of ignored associations between multiple interests. A source-domain data augmentation strategy was introduced, generating a denoised auxiliary sequence for each user in the source domain. Then the sparsity of user interaction data in the source domain was alleviated, and enriched user interest preferences were obtained. The interest extraction and multi-interest refinement transfer were implemented by utilizing the dual sequence multi-interest extraction module and the multi-interest refinement transfer module. Three publicly cross-domain recommendation evaluation tasks were conducted. The proposed model achieved the best performance compared with the best baseline, reducing the average MAE by 22.86% and the average RMSE by 19.65%, which verified the effectiveness of the method.

Key words： cold-start problem      cross-domain recommendation      data augmentation      multi-interest extraction      multi-interest refinement transfer     
Received: 19 November 2023      Published: 23 July 2024
CLC:  TP 391  
Fund:  国家自然科学基金资助项目 (62141201)；重庆市自然科学基金资助项目 (CSTB2022NSCQ-MSX1672)；重庆市教育委员会科学技术研究计划资助项目 (KJZD-M202201102)；重庆理工大学研究生教育高质量发展行动计划资助项目(gzlcx20233203).
Corresponding Authors: Xiaofei ZHU     E-mail: yinyabo@stu.cqut.edu.cn;zxf@cqut.edu.cn
Service
E-mail this article
Add to my bookshelf
Add to citation manager
E-mail Alert
RSS
Articles by authors
Yabo YIN
Xiaofei ZHU
Yidan LIU
Cite this article:

Yabo YIN,Xiaofei ZHU,Yidan LIU. Cross-domain recommendation model based on source domain data augmentation and multi-interest refinement transfer. Journal of ZheJiang University (Engineering Science), 2024, 58(8): 1717-1727.

URL:

https://www.zjujournals.com/eng/10.3785/j.issn.1008-973X.2024.08.018     OR     https://www.zjujournals.com/eng/Y2024/V58/I8/1717


源域数据增强与多兴趣细化迁移的跨域推荐模型

针对跨域推荐任务中源域用户交互数据不丰富所导致的兴趣偏好建模困难问题，以及多个兴趣之间的关联被忽略问题，提出源域数据增强与多兴趣细化迁移的跨域推荐模型. 该模型引入源域数据增强策略，为每个用户在源域中生成经过去噪处理的辅助序列，缓解用户在源域中的交互数据稀疏问题，获得更丰富的用户兴趣偏好. 使用双序列多兴趣提取模块和多兴趣细化迁移模块，完成兴趣提取与多个兴趣的细化迁移. 在基于3个公开跨域推荐评测任务的对比实验中，与最优的基线相比，提出方法的平均MAE降低了22.86%，平均RMSE降低了19.65%，取得了最优的性能表现，证明了提出方法的有效性.


关键词： 冷启动问题,  跨域推荐,  数据增强,  多兴趣提取,  多兴趣细化迁移 
Fig.1 Overall framework of CDR-ART model
Fig.2 Single domain pre-training and user auxiliary sequence generation
任务领域用户数量商品数量交互评分数量
源域目标域源域目标域重叠源域目标域源域目标域
任务1MovieMusic123 96075 25818 03150 05264 4431 697 5331 097 592
任务2BookMovie603 668123 96037 388367 98250 0528 898 0411 697 533
任务3BookMusic603 66875 25816 738367 98264 4438 898 0411 097 592
Tab.1 Task definition and data statistics
任务模型MAERMSE
β=20%β=50%β=80%β=20%β=50%β=80%
任务1TGT4.480 34.498 94.502 05.158 05.173 65.189 1
CMF1.520 91.689 32.418 62.015 82.227 13.093 6
DCDCSR1.491 81.814 42.719 41.921 02.343 93.306 5
SSCDR1.301 71.376 21.504 61.657 91.747 71.922 9
EMCDR1.235 01.327 71.500 81.551 51.664 41.877 1
PTUPCDR1.150 41.280 41.404 91.519 51.638 01.823 4
CDR-ART0.812 30.958 31.176 71.160 01.337 21.545 7
任务2TGT4.183 14.228 84.212 34.753 64.792 04.814 9
CMF1.363 21.581 32.157 71.791 82.088 62.677 7
DCDCSR1.397 11.673 12.361 81.734 62.055 12.770 2
SSCDR1.239 01.213 71.317 21.652 61.560 21.702 4
EMCDR1.116 21.183 21.315 61.412 01.498 11.643 3
PTUPCDR0.997 01.089 41.199 91.331 71.439 51.591 6
CDR-ART0.874 40.922 70.997 21.159 11.241 21.322 6
任务3TGT4.487 34.507 34.620 45.167 25.172 75.230 8
CMF1.828 42.128 23.013 01.382 92.727 53.694 8
DCDCSR1.841 12.173 63.140 52.295 52.677 13.584 2
SSCDR1.541 41.473 91.641 41.928 31.844 12.140 3
EMCDR1.352 41.473 21.719 11.673 71.800 02.111 9
PTUPCDR1.228 61.376 41.578 41.608 51.744 72.051 0
CDR-ART0.869 70.978 71.131 21.244 31.324 01.515 5
Tab.2 Overall performance of model on three tasks
任务$ \beta $/%MAE
CDR-ARTw/o auxw/o refinew/o aux&&refine
任务1200.812 30.847 00.865 40.943 5
500.958 31.048 71.041 31.084 3
801.176 71.255 71.255 91.342 9
任务2200.874 40.887 10.897 20.920 2
500.922 70.935 70.938 00.961 8
800.997 21.027 41.026 41.044 3
任务3200.869 70.909 90.944 50.972 8
500.978 71.040 81.038 71.080 9
801.131 21.189 71.195 81.245 9
Tab.3 Ablation experiment of CDR-ART on three tasks
$ M $任务1任务2任务3
MAERMSEMAERMSEMAERMSE
20.819 01.166 90.881 11.161 70.869 71.256 6
40.822 31.160 00.884 01.159 60.885 61.267 7
60.817 41.163 40.874 41.160 50.882 81.263 3
80.823 41.162 70.886 91.163 50.887 91.244 3
100.812 31.164 50.881 91.159 80.880 61.249 1
Tab.4 Impact of number of interests $ M $ on performance of model
Fig.3 Influence of number of potential interactive item on performance
Fig.4 Experiments on generalizability of CDR-ART
Fig.5 Performance of CDR-ART(full) and CDR-ART(w/o aux) for users with different lengths of interaction sequences in source domain
Fig.6 Case study of sampled user’s partial interaction with items
[1]   KOREN Y, BELL R, VOLINSKY C Matrix factorization techniques for recommender systems[J]. Computer, 2009, 42 (8): 30- 37
doi: 10.1109/MC.2009.263
[2]   HE X, LIAO L, ZHANG H, et al. Neural collaborative filtering [C]// Proceedings of the 26th International Conference on World Wide Web . Perth: ACM, 2017: 173-182.
[3]   LIU L, WANG L, LIAN T CaSe4SR: using category sequence graph to augment session-based recommendation[J]. Knowledge-Based Systems, 2021, 212 (1): 106558
[4]   QIAO J, WANG L Modeling user micro-behaviors and original interest via adaptive multi-attention network for session-based recommendation[J]. Knowledge-Based Systems, 2022, 244: 108567
doi: 10.1016/j.knosys.2022.108567
[5]   FU W, PENG Z, WANG S, et al. Deeply fusing reviews and contents for cold start users in cross-domain recommendation systems [C]// Proceedings of the AAAI Conference on Artificial Intelligence. Honolulu: AAAI, 2019: 94-101.
[6]   MAN T, SHEN H, JIN X, et al. Cross-domain recommendation: an embedding and mapping approach [C]// Proceedings of the 26th International Joint Conference on Artificial Intelligence. Melbourne: Morgan Kaufmann, 2017: 2464-2470.
[7]   KANG S K, HWANG J, LEE D, et al. Semi-supervised learning for cross-domain recommendation to cold-start users [C]// Proceedings of the 28th ACM International Conference on Information and Knowledge Management. Beijing: ACM, 2019: 1563-1572.
[8]   ZHU F, WANG Y, CHEN C, et al. A deep framework for cross-domain and cross-system recommendations [C]// Proceedings of the 27th International Joint Conference on Artificial Intelligence. Stockholm: Morgan Kaufmann, 2018: 3662-4402.
[9]   ZHU Y, TANG Z, LIU Y, et al. Personalized transfer of user preferences for cross-domain recommendation [C]// Proceedings of the 15th ACM International Conference on Web Search and Data Mining. [S. l. ]: ACM, 2022: 1507-1515.
[10]   SINGH A P, GORDON G J. Relational learning via collective matrix factorization [C]// Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Las Vegas: ACM, 2008: 650-658.
[11]   PAN W, XIANG E, LIU N, et al. Transfer learning in collaborative filtering for sparsity reduction [C]// Proceedings of the AAAI conference on artificial intelligence. Atlanta: AAAI, 2010: 230-235.
[12]   CAO J, CONG X, LIU T, et al. Item similarity mining for multi-market recommendation [C]// Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. Madrid: ACM, 2022: 2249-2254.
[13]   LI P, TUZHILIN A. Ddtcdr: deep dual transfer cross domain recommendation [C]// Proceedings of the 13th International Conference on Web Search and Data Mining. Houston: ACM, 2020: 331-339.
[14]   ZHAO C, ZHAO H, HE M, et al. Cross-domain recommendation via user interest alignment [C]// Proceedings of the ACM Web Conference. Austin: ACM, 2023: 887-896.
[15]   XIE R, LIU Q, WANG L, et al. Contrastive cross-domain recommendation in matching [C]// Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. Washington: ACM, 2022: 4226-4236.
[16]   CAO J, LIN X, CONG X, et al. Disencdr: learning disentangled representations for cross-domain recommendation [C]// Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. Madrid: ACM, 2022: 267-277.
[17]   LI C, XIE Y, YU C, et al. One for all, all for one: learning and transferring user embeddings for cross-domain recommendation [C]// Proceedings of the 16th ACM International Conference on Web Search and Data Mining. Singapore: ACM, 2023: 366-374.
[18]   LI C, LIU Z, WU M, et al. Multi-interest network with dynamic routing for recommendation at Tmall [C]// Proceedings of the 28th ACM International Conference on Information and Knowledge Management. Beijing: ACM, 2019: 2615-2623.
[19]   SABOUR S, FROSST N, HINTON G E. Dynamic routing between capsules [C]// Proceedings of the 31st International Conference on Neural Information Processing Systems . Long Beach: Curran Associates, 2017: 3859–3869.
[20]   CEN Y, ZHANG J, ZOU X, et al. Controllable multi-interest framework for recommendation [C]// Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. [S. l. ]: ACM, 2020: 2942-2951.
[21]   CHAI Z, CHEN Z, LI C, et al. User-aware multi-interest learning for candidate matching in recommenders [C]// Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. Madrid: ACM, 2022: 1326-1335.
[22]   LI J, ZHU J, BI Q, et al. MINER: multi-interest matching network for news recommendation [C]// Findings of the Association for Computational Linguistics. Dublin: ACL, 2022: 343-352.
[23]   WANG H, ZHANG F, XIE X, et al. DKN: deep knowledge-aware network for news recommendation [C]// Proceedings of the 2018 World Wide Web Conference. Lyon: ACM, 2018: 1835-1844.
[24]   FAN Z, XU K, ZHANG D, et al. Graph collaborative signals denoising and augmentation for recommendation [C]// Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. Taipei: ACM, 2023: 2037-2041.
[25]   WANG C, WANG Z, LIU Y, et al. Target interest distillation for multi-interest recommendation [C]// Proceedings of the 31st ACM International Conference on Information and Knowledge Management. Atlanta: ACM, 2022: 2007-2016.
[1] Tun YANG,Yongcun GUO,Shuang WANG,Xin MA. Obstacle recognition of unmanned rail electric locomotive in underground coal mine[J]. Journal of ZheJiang University (Engineering Science), 2024, 58(1): 29-39.
[2] Qing-lin AI,Jing-rui CUI,Bing-hai LV,Tong TONG. Surface defect detection method for bearing drum-shaped rollers based on fusion transformation of defective area[J]. Journal of ZheJiang University (Engineering Science), 2023, 57(5): 1009-1020.
[3] Hao-can XU,Ji-tuo LI,Guo-dong LU. Reconstruction of three-dimensional human bodies from single image by LeNet-5[J]. Journal of ZheJiang University (Engineering Science), 2021, 55(1): 153-161.