Scholarlink

<p>As education increasingly shifts toward online learning, there is a growing need for practical measures to support students and prevent them from falling behind. This need is particularly urgent given the high rates of failure and dropout that threaten educational outcomes. The challenge lies in the early identification and support of students at risk of underperforming, a task complicated by limited teacher resources. Technology offers a promising solution, particularly through the use of clickstream data to monitor students’ online activities. Clickstream data, which captures every click a student makes in a virtual learning environment, provides valuable insights into their engagement and learning patterns. These patterns vary across students and evolve over time, reflecting the complexity of the learning process. This research introduces a novel approach to predicting student performance by integrating advanced machine learning algorithms with clickstream data. We meticulously extract and organize data to facilitate a comprehensive analysis using both traditional techniques (such as Logistic Regression, Naive Bayes, K-Nearest Neighbors, Random Forest, and XGBoost) and sophisticated time-series models (including LSTM-XGBoost and GRU). Our findings reveal that the GRU algorithm, in particular, outperforms six established baseline methods, achieving an accuracy of <b>90.13%</b>. This advancement provides educators with new tools for delivering timely and personalized interventions, ultimately reducing failure and dropout rates and enhancing the overall quality of education.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

New Approach to Enhancing Student Performance Prediction Using Machine Learning Techniques and Clickstream Data in Virtual Learning Environments

  • 期刊论文

  • 发表日期：

    2025年2月5日

查看全文
  • Zakaria Khoudi,
  • Nasereddine Hafidi,
  • Mourad Nachaoui,
  • Soufiane Lyaqini

摘要

As education increasingly shifts toward online learning, there is a growing need for practical measures to support students and prevent them from falling behind. This need is particularly urgent given the high rates of failure and dropout that threaten educational outcomes. The challenge lies in the early identification and support of students at risk of underperforming, a task complicated by limited teacher resources. Technology offers a promising solution, particularly through the use of clickstream data to monitor students’ online activities. Clickstream data, which captures every click a student makes in a virtual learning environment, provides valuable insights into their engagement and learning patterns. These patterns vary across students and evolve over time, reflecting the complexity of the learning process. This research introduces a novel approach to predicting student performance by integrating advanced machine learning algorithms with clickstream data. We meticulously extract and organize data to facilitate a comprehensive analysis using both traditional techniques (such as Logistic Regression, Naive Bayes, K-Nearest Neighbors, Random Forest, and XGBoost) and sophisticated time-series models (including LSTM-XGBoost and GRU). Our findings reveal that the GRU algorithm, in particular, outperforms six established baseline methods, achieving an accuracy of 90.13%. This advancement provides educators with new tools for delivering timely and personalized interventions, ultimately reducing failure and dropout rates and enhancing the overall quality of education.