Regression-Based Student Performance Prediction Using the Grades of English, Math, and Programming Courses

One of the important roles of educators and academicians is to accurately assess students, and provide timely feedback to ensure that the learning outcomes of the academic program are achieved. Student performance prediction can help in this regard by predicting the future performance of the student based on the student's past performance benefitting from the data availability in the learning management system of the academic institution. This work aims to implement a machine learning model to predict the performance of students; particularly, students in the Computer Science and Engineering majors, based on their past performance in general first-year courses, namely, English, Math, and Programming. These particular courses were selected for the prediction purpose because they are essential in the academic progress of students in the Computer Science and Engineering majors. Specifically, English is the study language, yet it is considered a second language for many students, which can pose challenges in perceiving the core technical courses. Besides, Math and Programming courses are essential to understand the technical aspects of most Computer Science and Engineering courses. Therefore, knowing how the student is performing in these first-year courses will give important insights on how the student will perform later in the study program. In this work, the student performance is indicated by the Cumulative Grade Point Average (CGPA) which is predicted using polynomial regression based on a real dataset. The utilized dataset includes information about the CGPA and the grades of English, Math, and Programming courses from different levels (i.e., fresh and senior students), and extracted from the learning and management system of an academic institution. It is worth to mention that different degrees of the polynomial regression were investigated to observe how well the curve fits the data. To evaluate the prediction performance, the R-squared value is used. Results show an R-squared value of about 0.85 when the average of English, Math, and Programming courses is used, compared to a lower performance when the average of a combination of any two of these courses is used, and an even lower performance when only one course is used. These results provide two important conclusions, first, student performance prediction is an achievable goal in smart education systems, and second, the English, Math, and Programming courses can be used to predict the future performance of students specifically in the Computer Science and Engineering majors.