An End-to-End Personalized Preference Drift Aware Sequential Recommender System with Optimal Item Utilization

The user preference is dynamic and requires drift detection to capture changes for delivering relevant recommendations. A sequential recommender system with was proposed, where points are indicated by comparing characteristics of consecutive items. model leverages retrieve only interactions preferences the current preference. Nonetheless, number utilized items pre-defined may not be optimal. It also a unified architecture that optimizing each part individually. Recently, Content-Based Transformer has been proposed consider similar leveraging similarity function. trained in an end-to-end approach can applied recommendation task, as point item’s group changes. However, provides hard label, ignoring item real-world scenario exist many groups. For instance, most movies have multiple genres. This work proposes detects personalized pattern groups soft labels utilizes optimal amount jointly optimize deliver We conducted experiments verify effectiveness method it Transformers related methods. evaluation results show consistently outperforms baselines.