影响大学生学业成绩的多维因素探究——基于可解释的机器学习模型

A study of multi-factors affecting college students’learning performance——Based on interpretable machine learning model

在机器学习模型基础上加入SHAP解释方法,对影响大学生学业成绩的潜在因素进行深入研究,兼顾实际情境中预测准确性和指标可解释的要求。研究结果表明,大学生的个人因素和学校教育因素对其学业成绩影响较大,家庭因素对其学业成绩影响相对较小。从基于GBDT模型的SHAP结果可以看出,大学生的性别、年龄、是否有额外工作、是否有伴侣、上课是否听讲对其学业成绩都具有显著的影响。学校应从疏导大学生克服失败焦虑、引导其树立正确的恋爱观、帮助其树立终身学习理念、关注其实践创新能力四个维度对大学生提前进行干预,有针对性地提升其学业成绩。

This paper added SHAP interpretation method on the basis of machine learning model to conduct an in-depth study on the potential factors, which include the requirements of prediction accuracy and interpretation in the actual situation. The results show that personal factors and school education factors of college students have great influence, while family factors have relatively little influence . It can be seen that gender, age, extra work, partner, and lectures have a significant impact from the results of SHAP based on GBDT model. It is necessary for the university to intervene in advance from four dimensions: guiding the failure anxiety of the university students, guiding the university students to establish the correct view of love, helping the university students to establish the concept of lifelong learning and paying attention to the practical and innovative ability of the university students so as to improve the academic performance of the university students.