自 2020 年初新型冠状病毒的爆发以来，为了避免学生学习进度受到影响，学校的教学模式从线下转变为线上，这一转变不仅改变了教育的形式，也影响了整个教育过程，给学生和教师带来了一些挑战。本研究针对过往对学生群体研究受限于样本规模和数据维度的问题，在班杜拉的社会学习理论框架下，整合环境、行为、个人特征维度下的因素，通过对 2018 年和 2021 年全国基础教育质量监测 4 年级的数学监测数据的分析和挖掘，使用随机森林算法和 Lasso 回归总结成绩的影响因素的变化，使用聚类算法探究不同学业表现的学生亚群的变化，绘制出不同学业表现的小学生教育画像，从而探讨疫情如何影响学生的学业表现。主要研究结果如下： 通过随机森林和 Lasso 回归的筛选，比较疫情前后两年学业表现的影响因素，可以发现 2018 年即疫情前独有的影响因素有：教师行为、学习兴趣、课堂氛围和数学作业时间；2021 年即疫情后独有的影响因素有：学习自信心、家庭学习资源和学习动机。环境特征尤其是学校环境在疫情前对学生学业表现的影响更大，而家庭环境和个人特征在疫情后对学业表现的影响更大。 疫情前后不同学业表现的亚群体发生了变化，疫情前学业表现欠佳的学生有 3 类：高感受高状态型、低感受低兴趣型和高焦虑刻苦型，疫情后学业表现欠佳的学生变成 4 类：高感受高状态高条件型、低感受低状态型、高焦虑低条件型和刻苦型，家庭环境因素的影响作用变大。疫情前和疫情后学业表现优秀的学生分为 3 类，均为高条件型、低条件型和低感受低状态刻苦型，但是学习机会的影响在疫情后更加明显。不论学业表现如何，疫情后高感受高状态的学生都减少了，疫情对于学生的学校感受和学习状态存在负面影响。 本研究得出疫情前后学生学业表现的影响因素存在差异，不同学业水平的学生亚群发生变化，体现疫情对学生学业表现的影响，并针对不同亚群的学生提供更有针对性的指导，为改善在线学习体验、提高学生在线学习效果、推进个性化教学发展提供支持。

Since the outbreak of the COVID-19 in early 2020, schools have shifted their teaching mode from offline to online to avoid students' learning progress being affected, a shift that has not only changed the form of education but also affected the entire educational process, posing several challenges to students and teachers. This study addresses the problem that past studies on student groups are limited by sample size and data dimensions, and integrates factors under the environmental, behavioral, and personal characteristics dimensions under the framework of Bandura's social learning theory, and summarizes the factors influencing achievement using random forest algorithm and Lasso regression through the analysis and mining of mathematics monitoring data for grade 4 of the National Basic Education Quality Monitoring in 2018 and 2021 changes, using a clustering algorithm to explore changes in subgroups of students with different academic performance, and drawing an educational portrait of elementary school students with different academic performance, so as to explore how the epidemic affects students' academic performance. The main findings are as follows: By comparing the influences on academic performance in the two years before and after the epidemic through random forest and Lasso regression screening, it can be found that the influences unique to 2018, before the epidemic, are: teacher behavior, interest in learning, school climate, and time spent on math homework; and the influences unique to 2021, after the epidemic, are: learning self-confidence, home learning resources, and learning motivation. Environmental characteristics, especially school environment, had a greater impact on students' academic performance before the epidemic, while home environment and personal characteristics had a greater impact on academic performance after the epidemic. The subgroups with different academic performance changed before and after the epidemic. There were three categories of students who underperformed academically before the epidemic: high-feeling high-status, low-feeling low-interest, and high-anxiety hardworking, while students who underperformed academically after the epidemic became four categories: high-feeling highstatus high-conditioned, low-feeling low-status, high-anxiety low-conditioned, and hardworking, the influence of family environmental factors became more significant. Students who performed well academically before and after the epidemic were divided into 3 categories, all of which were high conditioned, low conditioned, and low feeling low state hardworking, but the effect of learning opportunities became more pronounced after the epidemic. Regardless of academic performance, there were fewer students with high school perception and high learning status after the epidemic, which means the epidemic had a negative impact on students' school perception and learning status. This study concludes that there are differences in the factors influencing students' academic performance before and after the epidemic, and changes in subgroups of students at different academic levels, which help understand the impact of the epidemic on students' academic performance and provide more targeted assistance for students at different academic levels, and provide support for improving the online learning experience, enhancing students' online learning effectiveness, and promoting the development of personalized instruction.