学生行为数据与学业成绩的关系研究——基于离群点检测算法

A correlational study on the student behavior data and academic performance

以S高校的校园一卡通数据为基础,通过对数据进行预处理并构建预测指标体系,使用基于聚类的离群点检测算法从多个角度对学生校园行为数据进行挖掘,并对学生行为与学业成绩的相关性进行分析.结果表明,每天用餐是否规律、进入图书馆的次数、借阅图书的次数、吃早饭次数之和与学业成绩具有强相关性.基于决策树算法构建了学业成绩预测模型,实现了对学生学业成绩预警.建议教学部门或教育管理者除了关注学生上课情况和学科成绩外,还应积极引导学生养成健康、规律、热爱阅读的生活方式.

Based on the campus one-card data of S university,a prediction index system is constructed by preprocessing the data.A clustering-based outlier detection algorithm is used to mine student campus behavior data from multiple perspectives,and the correlation between student behavior and academic performance is analyzed.Results indicate that there is a strong correlation among regularity of daily meals,number of times one visits library,number of times one borrows books,sum of times one eats breakfast,and academic performances.An academic performance prediction model is hence constructed for early warning of academic performance.Teachers or educational managers should not only pay attention to students'class and academic performance,but also actively guide students to develop health and love of reading.