Title:Pattern-wise Transparent Sequential Recommendation

Abstract:A transparent decision-making process is essential for developing reliable and trustworthy recommender systems. For sequential recommendation, it means that the model can identify critical items asthe justifications for its recommendation results. However, achieving both model transparency and recommendation performance simultaneously is challenging, especially for models that take the entire sequence of items as input without screening. In this paper,we propose an interpretable framework (named PTSR) that enables a pattern-wise transparent decision-making process. It breaks the sequence of items into multi-level patterns that serve as atomic units for the entire recommendation process. The contribution of each pattern to the outcome is quantified in the probability space. With a carefully designed pattern weighting correction, the pattern contribution can be learned in the absence of ground-truth critical patterns. The final recommended items are those items that most critical patterns strongly endorse. Extensive experiments on four public datasets demonstrate remarkable recommendation performance, while case studies validate the model transparency. Our code is available at this https URL.