Exploring Student Performance Using Machine Learning: Effects of Attendance, Weekly Quizzes, and Midterm Exams
Keywords:
students’ performance, educational data mining, K-Means clustering, Random Forest, SHAP analysis.Abstract
This research aims to explore the most influential factor of Absence, Weekly-quizzes and Midterm-Exam on Final-Exam score of 164 female student enrolled in first semester in fundamental information technology course. Using descriptive statistics and Pearson correlation analysis, K-Means clustering and regression models (Decision Tree, Random Forest, Gradient Boosting).
The Pearson correlation results show the strongest correlation was between Midterm-Exam and Final-Exam (r=0.667), followed by Weekly-Quizzes (r=0.612), while Absence had low negative correlation with (r== –0.359).
K-Means clustering revealed variation in students’ performance and were classified into three groups high, average and low performance, while Random Forest achieved the most accurate prediction with (R²=0.62) and least prediction errors.
SHAP analysis confirmed that Midterm-Exam was most influential factor on Final-Exam score, followed with Weekly-Quizzes, while Absence had indirect effect on Final-Exam score. In conclusion, this study confirms the importance of continuous assessment to predict students at risk and providing appropriate support in the right time to enhance the overall education results.
Downloads
