Scholarlink

Student dropout is a widespread problem in higher education that has significant consequences for both individuals and institutions. This study aims to explore the predictive capabilities of various machine learning models in identifying students who are at risk of dropping out. The analysis includes traditional models such as Decision Trees (DT), Naive Bayes (NB), and Support Vector Machines (SVM), as well as advanced techniques like TabNet and DatRet. Ensemble learning methods, including Boosting, Bagging, and Stacking, are also incorporated to increase predictive accuracy. In addition to individual model performance, this research introduces a novel approach called sequential modelling. Here, the outputs of DT, NB, and SVM are sequentially integrated to improve predictive accuracy. Combining the outputs of these models helps to identify dropout patterns more effectively than relying on any one model alone. Notably, the simpler machine learning models sequentially performed better than TabNet and DatRet. The findings of this study are important for the ongoing discussion on how to improve student retention rates. A diverse and well-informed approach is necessary to address this complex challenge. Predicting dropout patterns is essential for implementing timely interventions and support mechanisms, which can create a positive learning environment for all students.

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

Student Dropout Prediction Using Sequential Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shriya Goswami,
  • Sufi Farhan Murshed,
  • Akash Chandra

摘要

Student dropout is a widespread problem in higher education that has significant consequences for both individuals and institutions. This study aims to explore the predictive capabilities of various machine learning models in identifying students who are at risk of dropping out. The analysis includes traditional models such as Decision Trees (DT), Naive Bayes (NB), and Support Vector Machines (SVM), as well as advanced techniques like TabNet and DatRet. Ensemble learning methods, including Boosting, Bagging, and Stacking, are also incorporated to increase predictive accuracy. In addition to individual model performance, this research introduces a novel approach called sequential modelling. Here, the outputs of DT, NB, and SVM are sequentially integrated to improve predictive accuracy. Combining the outputs of these models helps to identify dropout patterns more effectively than relying on any one model alone. Notably, the simpler machine learning models sequentially performed better than TabNet and DatRet. The findings of this study are important for the ongoing discussion on how to improve student retention rates. A diverse and well-informed approach is necessary to address this complex challenge. Predicting dropout patterns is essential for implementing timely interventions and support mechanisms, which can create a positive learning environment for all students.