Candidate-dependent routing helps recall but lowers top accuracy in recommendations
Candidate-dependent residual routing trades top-of-list accuracy for recall in sequential recommendation
Information RetrievalMachine Learning
Summary
Recommender systems try to guess which items a user will like based on their past behavior. This paper studies a method that changes how much each past action counts depending on the item being recommended. The authors found this method makes the system find more relevant items overall but pushes the very best matches lower in the ranking. So it finds more good options but is less accurate at picking the single best one.
What this means in practice
- •For ecommerce platform developers: Improve product recommendation recall to show more relevant items in the top ten for users with recent shopping history.
- •For digital marketing teams: Tailor campaign targeting by identifying more relevant products further down in recommendation lists for users with recent interactions.
Tested on one dataset.
Authors
Han Chen, Yingrui Li
Abstract
Candidate-dependent routing lets a sequential recommender weight a user's past behaviors differently for each item it scores. We test whether this flexibility helps by adding a small residual routing module to a self-attentive baseline, in two versions that differ only in what sets the routing weights: the user's history representation or the candidate item's embedding. Across five paired seeds on the REES46 cosmetics-shop dataset, with matched training for all three models, candidate-dependent routing leaves the primary metric, normalized discounted cumulative gain at cutoff ten (NDCG@10), essentially unchanged (mean relative change +0.003%). Secondary and exploratory analyses show that this flat average hides two effects that cancel. The candidate-dependent model places more relevant items in its top ten (recall +1.81%) but ranks the first one lower (mean reciprocal rank -1.09%), and most of its new hits land at positions six to ten. Its gains come from queries with at least one target among the user's last 50 events, while queries whose targets all lie outside that window get worse. In this form, candidate-dependent routing trades accuracy at the top of the list for recall further down.