预测学生在大规模开放在线课程中的学习成果：基于机器学习模型的综合分析
Predicting student success in MOOCs: a comprehensive analysis using machine learning models.
作者信息
Althibyani Hosam A
机构信息
Learning Design and Technology Department, College of Education, University of Jeddah, Jeddah, Saudi Arabia.
出版信息
PeerJ Comput Sci. 2024 Aug 23;10:e2221. doi: 10.7717/peerj-cs.2221. eCollection 2024.
BACKGROUND
This study was motivated by the increasing popularity of Massive Open Online Courses (MOOCs) and the challenges they face, such as high dropout and failure rates. The existing knowledge primarily focused on predicting student dropout, but this study aimed to go beyond that by predicting both student dropout and course results. By using machine learning models and analyzing various data sources, the study sought to improve our understanding of factors influencing student success in MOOCs.
OBJECTIVES
The primary aim of this research was to develop accurate predictions of students' course outcomes in MOOCs, specifically whether they would pass or fail. Unlike previous studies, this study took into account demographic, assessment, and student interaction data to provide comprehensive predictions.
METHODS
The study utilized demographic, assessment, and student interaction data to develop predictive models. Two machine learning methods, logistic regression, and random forest classification were employed to predict students' course outcomes. The accuracy of the models was evaluated based on four-class classification (predicting four possible outcomes) and two-class classification (predicting pass or fail).
RESULTS AND CONCLUSIONS
The study found that simple indicators, such as a student's activity level on a given day, could be as effective as more complex data combinations or personal information in predicting student success. The logistic regression model achieved an accuracy of 72.1% for four-class classification and 92.4% for 2-class classification, while the random forest classifier achieved an accuracy of 74.6% for four-class classification and 95.7% for two-class classification. These findings highlight the potential of machine learning models in predicting and understanding students' course outcomes in MOOCs, offering valuable insights for improving student engagement and success in online learning environments.
背景
本研究的动机源于大规模在线开放课程（MOOCs）日益普及及其面临的挑战，如高辍学率和不及格率。现有知识主要集中于预测学生辍学情况，但本研究旨在超越这一点，通过预测学生辍学情况和课程成绩来进行研究。通过使用机器学习模型并分析各种数据源，该研究试图增进我们对影响MOOCs中学生成功因素的理解。
目标
本研究的主要目的是准确预测MOOCs中学生的课程成绩，特别是他们是否会通过或不及格。与以往研究不同，本研究考虑了人口统计学、评估和学生互动数据，以提供全面的预测。
方法
该研究利用人口统计学、评估和学生互动数据来开发预测模型。采用逻辑回归和随机森林分类这两种机器学习方法来预测学生的课程成绩。基于四类分类（预测四种可能结果）和两类分类（预测通过或不及格）对模型的准确性进行评估。
结果与结论
该研究发现，简单指标，如学生在某一天的活动水平，在预测学生成功方面可能与更复杂的数据组合或个人信息一样有效。逻辑回归模型在四类分类中的准确率为72.1%，在两类分类中的准确率为92.4%，而随机森林分类器在四类分类中的准确率为74.6%，在两类分类中的准确率为95.7%。这些发现凸显了机器学习模型在预测和理解MOOCs中学生课程成绩方面的潜力，为提高在线学习环境中学生的参与度和成功率提供了有价值的见解。
Zotero 插件