CFSeRec: A Contrastive Framework for Sequential Recommendation

Tong Wang, Yaping Dai, Shuai Shao*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In sequential recommender systems, the main problems are the long-tailed distribution of data and noise interference. A Contrastive Framework for Sequential Recommendation (CFSeRec) is proposed to solve these two problems respectively. Token shuffling and adversarial attack data augmentation methods are used in the framework to improve the quality and quantity of training data, so that the long-tailed problem is mitigated. Through the application of projection head method, the sequence representation becomes more general and robust, rather than just adapted to the task of contrastive learning. Therefore, the impact of noise on sequence recommender systems is effectively alleviated. Experiments on four public datasets show that CFSeRec achieves state-of-the-art performance in the metrics of hit ratio and normalized discounted cumulative gain, when comparing to the seven previous frameworks.

Original languageEnglish
Title of host publication2023 42nd Chinese Control Conference, CCC 2023
PublisherIEEE Computer Society
Pages8211-8216
Number of pages6
ISBN (Electronic)9789887581543
DOIs
Publication statusPublished - 2023
Event42nd Chinese Control Conference, CCC 2023 - Tianjin, China
Duration: 24 Jul 202326 Jul 2023

Publication series

NameChinese Control Conference, CCC
Volume2023-July
ISSN (Print)1934-1768
ISSN (Electronic)2161-2927

Conference

Conference42nd Chinese Control Conference, CCC 2023
Country/TerritoryChina
CityTianjin
Period24/07/2326/07/23

Keywords

  • Attention Mechanism
  • Contrastive Learning
  • Self-supervised Learning
  • Sequential Recommendation

Fingerprint

Dive into the research topics of 'CFSeRec: A Contrastive Framework for Sequential Recommendation'. Together they form a unique fingerprint.

Cite this