Scholarlink

In recent decades, data mining in education has become increasingly significant due to the proliferation of accessible educational data. This study assesses the effectiveness of three machine learning algorithms—Logistic Regression, Decision Tree, and Random Forest —in forecasting academic outcomes using the Open University Learning Analytics Dataset (OULAD) [24], then compared with a Deep Learning Long Short-Term Memory (LSTM) model. The models underwent training and testing to enhance performance through hyperparameter optimization and cross-validation. Course FFF demonstrated the highest predictive accuracy among the evaluated courses, with Random Forest surpassing the other models by attaining 92% accuracy, accompanied by robust recall, precision, and F1-score. Decision Tree and Logistic Regression exhibited an accuracy of 83%, and 81%, while LSTM achieved 87%. The findings show the efficacy of ensemble learning methods, such as Random Forest, in discerning patterns in student performance and facilitating early intervention strategies. Future endeavors may augment model performance by further integrating supplementary features and optimizing hyperparameters to enhance predictive accuracy.

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

Machine Learning and Deep Learning Models for Early Identification of At-Risk Students Using OULAD VLE Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yousuf Nasser Al Husaini,
  • Mohammed Abdulla Al Husaini,
  • Wasin Ayman Al Kishri,
  • Mahmood Al Bahri,
  • Nur Syufiza Ahmad Shukor

摘要

In recent decades, data mining in education has become increasingly significant due to the proliferation of accessible educational data. This study assesses the effectiveness of three machine learning algorithms—Logistic Regression, Decision Tree, and Random Forest —in forecasting academic outcomes using the Open University Learning Analytics Dataset (OULAD) [24], then compared with a Deep Learning Long Short-Term Memory (LSTM) model. The models underwent training and testing to enhance performance through hyperparameter optimization and cross-validation. Course FFF demonstrated the highest predictive accuracy among the evaluated courses, with Random Forest surpassing the other models by attaining 92% accuracy, accompanied by robust recall, precision, and F1-score. Decision Tree and Logistic Regression exhibited an accuracy of 83%, and 81%, while LSTM achieved 87%. The findings show the efficacy of ensemble learning methods, such as Random Forest, in discerning patterns in student performance and facilitating early intervention strategies. Future endeavors may augment model performance by further integrating supplementary features and optimizing hyperparameters to enhance predictive accuracy.