Review sentiment classification aims to predict user sentiment for given user-generated review. Most of the existing methods enhance their sentiment classifiers by incorporating user information. However, those methods normally ignore user's mood, which could influence her/his sentiment expression. Actually, users in a certain mood tend to express mood-congruent sentiment. Furthermore, the related studies may not fully utilize user's personality to model her/his mood. In this paper, we are motivated to propose a Personality-Assisted Mood for Sentiment Classification (PAMSC) model to classify user sentiment. Concretely, we first adopt the target review with global user preference to prejudge user sentiment. Meanwhile, we model the personalized mood impact on her/his sentiment expression according to the corresponding historical reviews. We finally obtain the prediction result by utilizing the personalized mood impact to adjust the prejudged sentiment distribution. Particularly, personality plays two major roles in the process of mood modeling. One is to analyze the personalized duration of user's mood, and the other is to model the personalized attention toward mood-congruent information. The experimental results demonstrate that our PAMSC model not only achieves the highest classification accuracy than the related models on three real-world datasets but also has stronger interpretability for the prediction process. •Classify user sentiment with personalized mood impact based on the cognitive mechanism of mood-congruent attention bias.•Extract the user's mood from her/his historical reviews based on the findings of linguistic psychology.•Analyze the personalized duration of the user's mood based on the user's Big-Five personality.•Adopt user's Big-Five personality to model personalized attention towards mood-congruent information under different moods•Achieve better classification accuracy and interpretation than the related baseline models on three real-world datasets.