MOOCs学习行为与学习效果的逻辑回归分析

A logistic regression analysis of learning behaviors and learning outcomes in MOOCs

目前,MOOCs(大规模开放在线课程)在世界范围内迅猛发展,但是随之而来的是对MOOCs学习质量和高辍学率等现象的质疑。现有MOOCs平台大都对学习者在线学习行为有较为详细的记录。如何对学习行为数据进行分析、建模和解读是大数据时代教育研究的热点和难点所在。逻辑回归方法作为一种成熟的机器学习方法可以有效地建立学习行为和学习效果之间的模型。本研究总结了在线学习领域逻辑回归研究的流程,在此基础上,从MOOCs在线学习过程出发构建了学习行为指标,并应用逻辑回归对MOOCs学习数据进行分析,就学习者在线学习行为对学习成绩的影响展开了探索。研究检验了逻辑回归对于在线学习效果研究的价值,发现了课程注册时滞、登录课程次数、提交作业测试次数、习题保存次数的均值和视频观看完成度等指标与成绩的相关性。研究发现:在该课程中提交作业测试可以作为MOOCs学习成绩预测的关键指标,所构建的逻辑回归模型预测准确率达到98%。

The rapid growth of Massive Open Online Courses(MOOC) on a global scale has brought with them doubts about learning quality and concerns about their high attrition rates. Generally speaking, MOOC platforms keep a detailed record of learners' online learning behaviors. Analysis, modeling and interpretation of these data is top on the educational research agenda in the era of big data.Logistic regression analysis is used in this study to explore the impact of online learning behaviors on learning outcomes. Findings from the study indicate that learning outcomes are correlated with delay in course registration, how many times that learners log in and submit their assignments/quizzes, the average number of times they save their exercises, and the extent to which video clips are watched. It is also found that assignment/quiz submissions can serve as a key index in predicting MOOC learning outcomes and that the resultant logistic regression model has 98% prediction accuracy.