Contrastive Cross-Domain Sequential Recommendation

Abstract

Cross-Domain Sequential Recommendation (CDSR) aims to predict future\ninteractions based on user's historical sequential interactions from multiple\ndomains. Generally, a key challenge of CDSR is how to mine precise cross-domain\nuser preference based on the intra-sequence and inter-sequence item\ninteractions. Existing works first learn single-domain user preference only\nwith intra-sequence item interactions, and then build a transferring module to\nobtain cross-domain user preference. However, such a pipeline and implicit\nsolution can be severely limited by the bottleneck of the designed transferring\nmodule, and ignores to consider inter-sequence item relationships. In this\npaper, we propose C^2DSR to tackle the above problems to capture precise user\npreferences. The main idea is to simultaneously leverage the intra- and inter-\nsequence item relationships, and jointly learn the single- and cross- domain\nuser preferences. Specifically, we first utilize a graph neural network to mine\ninter-sequence item collaborative relationship, and then exploit sequential\nattentive encoder to capture intra-sequence item sequential relationship. Based\non them, we devise two different sequential training objectives to obtain user\nsingle-domain and cross-domain representations. Furthermore, we present a novel\ncontrastive cross-domain infomax objective to enhance the correlation between\nsingle- and cross- domain user representations by maximizing their mutual\ninformation. To validate the effectiveness of C^2DSR, we first re-split four\ne-comerce datasets, and then conduct extensive experiments to demonstrate the\neffectiveness of our approach C^2DSR.\n