Scholarlink

Higher education institutions need early student performance predictions to provide timely academic support. This research investigates the requirements for practical studies that use complete continuous assessment information in the Arab academic environment. This research aimed to create and assess machine learning (ML) prediction models for determining university students’ final numerical grades in their courses. The research data included grades from 65 students who completed three daily exams and three daily assignments and a college assignment and a report and a midterm exam. The study implemented a quantitative applied research approach that involved data preprocessing and 5-fold cross-validation of Linear Regression (LR) and Decision Tree (DT) Regressor and Support Vector Regressor (SVR) models. The evaluation criteria consisted of Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and R-squared (R2). The LR model produced the highest predictive accuracy, with an average R2 value of 0.829, an average MAE of 6.4874 and an average RMSE of 8.3493. The mid exam score emerged as the most important predictor according to feature analysis. This research adds value through its application of ML methods to actual assessment data from an Arab university, which reveals how different assessment elements impact student achievement. This research offers actionable suggestions for teachers to detect struggling students while helping them develop better assessment methods. Research needs to expand its dataset while investigating sophisticated algorithms for future development.

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

Early Prediction of Performance: A Machine Learning Approach to Predicting Student Performance Using Continuous Assessment Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ammar Farooq Abbas,
  • Yahya Layth Khaleel,
  • Mustafa Abdulfattah Habeeb,
  • Noor Walid Khalid,
  • Saba Hussein Rashid,
  • Noor Saud Abd

摘要

Higher education institutions need early student performance predictions to provide timely academic support. This research investigates the requirements for practical studies that use complete continuous assessment information in the Arab academic environment. This research aimed to create and assess machine learning (ML) prediction models for determining university students’ final numerical grades in their courses. The research data included grades from 65 students who completed three daily exams and three daily assignments and a college assignment and a report and a midterm exam. The study implemented a quantitative applied research approach that involved data preprocessing and 5-fold cross-validation of Linear Regression (LR) and Decision Tree (DT) Regressor and Support Vector Regressor (SVR) models. The evaluation criteria consisted of Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and R-squared (R2). The LR model produced the highest predictive accuracy, with an average R2 value of 0.829, an average MAE of 6.4874 and an average RMSE of 8.3493. The mid exam score emerged as the most important predictor according to feature analysis. This research adds value through its application of ML methods to actual assessment data from an Arab university, which reveals how different assessment elements impact student achievement. This research offers actionable suggestions for teachers to detect struggling students while helping them develop better assessment methods. Research needs to expand its dataset while investigating sophisticated algorithms for future development.