Effective Utilization of Large-scale Unobserved Data in Recommendation Systems

被引:0
作者
Zhang, Feng [1 ]
Xu, Yulin [1 ]
Chen, Hongjie [1 ]
Yuan, Xu [1 ]
Liu, QingWen [1 ]
Jiang, YuNing [2 ]
机构
[1] Taotian Grp, Hangzhou, Zhejiang, Peoples R China
[2] Taotian Grp, Beijing, Peoples R China
来源
PROCEEDINGS OF THE 33RD ACM INTERNATIONAL CONFERENCE ON INFORMATION AND KNOWLEDGE MANAGEMENT, CIKM 2024 | 2024年
关键词
Unobserved items; Recommender System; Transfer Learning; Multi-domain recommendation;
D O I
10.1145/3627673.3680067
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Ranking models play an important role in industrial recommendation systems. However, most ranking models are trained only with the observed items but used to retrieve all items in the entire space, which may suffer from the sample selection bias and the exposure bias. Inspired by the entire space learning framework, we carry out detailed data analyses on large-scale unobserved items and find that they contain quite a few "potentially-positive" samples. In this paper, we propose an "Extract and Transfer" (EAT) framework, utilizing quantities of unobserved items and other domains' data to construct more training data for ranking models. Specifically, we first extract "potentially-positive" samples and negative ones according to their ranking scores from the unobserved data, and then design an Entire Space Transfer Learning (ESTL) model to transfer knowledge between observed and unobserved samples, instead of directly mixing them together to avoid negative transfer. Experiments on production data collected from Taobao validate the proposed method's superiority. Besides, we have deployed EAT on the Taobao recommendation system, obtaining 6.22% IPV (Item Page View) and 3.77% CTR improvement. The code is available at https://github.com/Recommender1/EAT.git(1).
引用
收藏
页码:5070 / 5077
页数:8
相关论文
共 29 条
[1]   PEPNet: Parameter and Embedding Personalized Network for Infusing with Personalized Prior Information [J].
Chang, Jianxin ;
Zhang, Chenbin ;
Hui, Yiqun ;
Leng, Dewei ;
Niu, Yanan ;
Song, Yang ;
Gai, Kun .
PROCEEDINGS OF THE 29TH ACM SIGKDD CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING, KDD 2023, 2023, :3795-3804
[2]   Bias and Debias in Recommender System: A Survey and Future Directions [J].
Chen, Jiawei ;
Dong, Hande ;
Wang, Xiang ;
Feng, Fuli ;
Wang, Meng ;
Xiangnan, He .
ACM TRANSACTIONS ON INFORMATION SYSTEMS, 2023, 41 (03)
[3]  
Cheng H.-T., 2016, P 1 WORKSH DEEP LEAR, P7
[4]   Deep Neural Networks for YouTube Recommendations [J].
Covington, Paul ;
Adams, Jay ;
Sargin, Emre .
PROCEEDINGS OF THE 10TH ACM CONFERENCE ON RECOMMENDER SYSTEMS (RECSYS'16), 2016, :191-198
[5]   Neighborhood-based Hard Negative Mining for Sequential Recommendation [J].
Fan, Lu ;
Pu, Jiashu ;
Zhang, Rongsheng ;
Wu, Xiao-Ming .
PROCEEDINGS OF THE 46TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL, SIGIR 2023, 2023, :2042-2046
[6]   Graph Neural Networks for Social Recommendation [J].
Fan, Wenqi ;
Ma, Yao ;
Li, Qing ;
He, Yuan ;
Zhao, Eric ;
Tang, Jiliang ;
Yin, Dawei .
WEB CONFERENCE 2019: PROCEEDINGS OF THE WORLD WIDE WEB CONFERENCE (WWW 2019), 2019, :417-426
[7]   Cross-domain Recommendation Without Sharing User-relevant Data [J].
Gao, Chen ;
Chen, Xiangning ;
Feng, Fuli ;
Zhao, Kai ;
He, Xiangnan ;
Li, Yong ;
Jin, Depeng .
WEB CONFERENCE 2019: PROCEEDINGS OF THE WORLD WIDE WEB CONFERENCE (WWW 2019), 2019, :491-502
[8]   Learning to Select Instance: Simultaneous Transfer Learning and Clustering [J].
Huan, Zhaoxin ;
Wang, Yulong ;
He, Yong ;
Zhang, Xiaolu ;
Fu, Chilin ;
Wu, Weichang ;
Zhou, Jun ;
Ding, Ke ;
Zhang, Liang ;
Mo, Linjian .
SIGIR '21 - PROCEEDINGS OF THE 44TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL, 2021, :1950-1954
[9]  
Jiang Jing, 2007, P ACL, P264
[10]  
Li Bin, 2009, Proceedings of the 26th Annual International Conference on Machine Learning, V382, P617, DOI 10.1145/1553374