A Robust Sequential Recommendation Model Based on Multiple Feedback Behavior Denoising and Trusted Neighbors

Abstract

At present, most of the personalized sequential recommendations utilize users’ implicit positive feedback (such as clicks) to predict user behavior, ignoring the impact of implicit negative feedback and explicit feedback on the accuracy of recommendation results prediction. In this paper, we propose a robust sequence recommendation model based on multi feedback behavior denoising and trusted neighbors, which utilizes multiple feedback behavior data for feature denoising and considers trusted nearest neighbor information to improve model performance. Firstly, by learning the feature representations and interactions of various types of feedback, explicit feedback is used to map and purify implicit feedback with the same and different attributes, resulting in unbiased user performance. Then, we design a filter attention network to identify highly trusted neighbor information. Finally, we integrate pure user interest representations and trusted nearest neighbor representations to improve the accuracy and robustness of the model. The experimental results on two publicly available datasets show that the proposed sequential recommendation model can achieve superior results to baseline methods in both AUC and RelaImpr.