Towards Open-World Cross-Domain Sequential Recommendation: A Model-Agnostic Contrastive Denoising Approach

Wujiang Xu, Xuying Ning, Wenfang Lin, Mingming Ha, Qiongxu Ma, Qianqiao Liang, Xuewen Tao, Linxun Chen, Bing Han, Minnan Luo

Abstract: Cross-domain sequential recommendation (CDSR) aims to address the data spCH). Recently, some SR approaches have utilized auxiliary behaviors to complement the information for long-tailed users. However, these methods cannot deliver promising performance in CDSR, as they overlook the semantic gap between target and auxiliary behaviors, as well as user interest deviation across domains (2nd CH). In this paper, we propose a model-agnostic contrastive denoising (MACD) approach towards open-world CDSR. We introduce auxiliary behavior sequence information (i.e., clicks) into CDSR methods to explore potential interests. Specifically, we design a denoising interest-aware network combined with a contrastive information regularizer to remove inherent noise from auxiliary behaviors and exploit multi-interest from users. Extensive offline experiments on public industry datasets and a standard A/B test on a large-scale financial platform with millions of users both confirm the remarkable performance of our model in open-world CDSR scenarios. Code and dataset are available at https://github.com/WujiangXu/MACD.