With the advancement of internet technology, responsibly recommending the most suitable items to users has emerged as a pivotal challenge for online e-commerce platforms. However, existing works fail to combine users’ sentiments with behavioral characteristics to provide effective and explainable recommendations. In this paper, we propose a novel Explainable Recommendation method, termed USB-ER, which integrates User Sentiment and Behavior, thereby enhancing the accuracy and explainability of the recommendation. Specifically, we first propose a sentiment classification model based on user reviews. Different from traditional classification models, our model utilizes a multi-task learning framework to extract fine-grained sentiments at both the rating and review levels, subsequently fusing them to ensure a precise quantification of the Positive Review Rate (PRR). Then we propose a PRR-based explainable recommendation model that integrates collaborative filtering with items’ PRR values, thereby enhancing the recommendation quality and enabling more personalized explanations. We conduct extensive experiments on real-world dataset. The results demonstrate the effectiveness of our proposed method.