TCA4Rec: Contrastive Learning with Popularity-aware Asymmetric Augmentation for Robust Sequential Recommendation DOI
Yanan Bai, Xiaolu Li, Chunming Xia

et al.

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: April 11, 2025

Abstract Sequential recommender systems play a pivotal role in modern recommendation scenarios by capturing users' dynamic interests through their historical interactions. While existing methods often rely on sophisticated deep models to enhance quality, they suffer from performance degradation due sparse supervision signals and popularity bias the training data. In this paper, we propose TCA4Rec, robust sequential framework that addresses these challenges via novel two-stage contrastive learning approach. Our incorporates an additional memory module aggregate sequence embeddings, thereby providing flexible generalized representations of user preferences. To mitigate bias, derive Asymmetric Multi-instance Noise Contrastive Estimation (AMINCE) loss function supplies rich, bias-aware signals, while our strategy significantly reduces over-dominance popular items during optimization. \added{Extensive experiments three real-world datasets demonstrate TCA4Rec achieves significant improvements over state-of-the-art baselines. Specifically, it attains absolute gains 19.26% HR@5 17.97% NDCG@5 Amazon-sports dataset. The also shows promising practical potential for applications e-commerce engines, video streaming platforms requiring long-tail content exposure, computational advertising where mitigating can directly impact advertiser ROI.}

Language: Английский

Transforming E-Commerce with Intelligent Recommendation Systems: A Review of Current Trends in Machine Learning and Deep Learning DOI Open Access
P. Chinnasamy

International Journal of Computational and Experimental Science and Engineering, Journal Year: 2025, Volume and Issue: 11(2)

Published: March 17, 2025

In the ever-changing realm of E-Commerce, it is essential for online businesses to comprehend and adjust shifting consumer behaviour in order achieve long-term success. which, Intelligent Recommendation System (IRS) has gained familiarity by suggesting personalized information based on user preference behaviours. Hence, review paper primarily aims analyse significance intelligent recommendation system transform ecommerce field, specifically enrich personalisation satisfaction, enhance revenue business. Accordingly, proposed survey discussed traditional AI-powered personalization ecommerce. utilize sophisticated algorithms extensive data, allowing provision highly customized relevant content, product recommendation, satisfaction. Besides, examines future trends AI integration within e-commerce, particularly advancements Natural Language Processing (NLP) visual search technologies, which are poised further The concludes with a look toward directions technologies anticipating NLP capabilities, promise shopping experience. Overall, findings article underscores transformative impact IRS e-commerce sector, advocating their continued development response evolving market demands.

Language: Английский

Citations

0
TCA4Rec: Contrastive Learning with Popularity-aware Asymmetric Augmentation for Robust Sequential Recommendation DOI
Yanan Bai, Xiaolu Li, Chunming Xia

et al.

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: April 11, 2025

Abstract Sequential recommender systems play a pivotal role in modern recommendation scenarios by capturing users' dynamic interests through their historical interactions. While existing methods often rely on sophisticated deep models to enhance quality, they suffer from performance degradation due sparse supervision signals and popularity bias the training data. In this paper, we propose TCA4Rec, robust sequential framework that addresses these challenges via novel two-stage contrastive learning approach. Our incorporates an additional memory module aggregate sequence embeddings, thereby providing flexible generalized representations of user preferences. To mitigate bias, derive Asymmetric Multi-instance Noise Contrastive Estimation (AMINCE) loss function supplies rich, bias-aware signals, while our strategy significantly reduces over-dominance popular items during optimization. \added{Extensive experiments three real-world datasets demonstrate TCA4Rec achieves significant improvements over state-of-the-art baselines. Specifically, it attains absolute gains 19.26% HR@5 17.97% NDCG@5 Amazon-sports dataset. The also shows promising practical potential for applications e-commerce engines, video streaming platforms requiring long-tail content exposure, computational advertising where mitigating can directly impact advertiser ROI.}

Language: Английский

Citations

0