.png){kind=icon}
.png){kind=icon}
Return
Proxy-enhanced cross-domain sequential recommendation
.png){kind=icon}
PRE
AI
DOI：10.1007/s10618-025-01097-4.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Cross-domain sequential recommendation (CDSR) aims to predict the next item that a user is most likely to interact with based on past sequential behavior from multiple domains. Existing works on CDSR usually transfer knowledge across different domains by linking items between domains via common users, which suffers from the following limitations: (1) Due to the inherent differences between the domains, transferring information across domains can be affected by different representations of related items in different domains. (2) User behavior in various domains may not always contribute constructively to the recommendation process in the target domain. (3) The overall semantic characterization of cross-domain sequences tends to be dominated by data-rich domains. In this work, we propose a novel cross-domain sequential recommendation model to address the above challenges. Specifically, we first develop a proxy encoder that uses textual descriptions to pre-train a universal representation for each item across all domains. We then design a feature enhancement module to filter out extraneous information from the textual descriptions, thereby optimizing and enriching this representation. Moreover, we present a contrastive learning auxiliary task to enhance a cross-domain sequence by weighing the importance of the items in the auxiliary domain with respect to the target domain. Finally, in the prediction phase, we construct the pre-trained prompt representations for each domain to balance the training of the two domains. Extensive experiments demonstrate the superiority of our proposed method from various aspects.
Keywords:
Cross-domain sequential recommendation
Contrastive learning
Universal representation
Journal
IF:
4.3
Papers:
192
Citations:
6.0K