A LINEAR REGRESSION MODEL FOR PREDICTING STUDENTS' FINAL TEST SCORES BASED ON ASSESSMENT RESULTS

In modern higher education, assessing students’ performance and predicting their future outcomes is becoming an essential part of educational analytics. This paper presents a linear regression model for predicting students’ final test scores based on their assessment results within a credit-modular system. The assessment process is divided into three main components: current (continuous) assessment, midterm assessment, and final examination. Since each component is assigned a maximum score (e.g., 30 for current, 20 for midterm, and 50 for final), the results are normalized into percentages to ensure consistency across different subjects and institutions. The proposed model demonstrates that using normalized percentages increases prediction accuracy and allows instructors to design personalized learning strategies. International research findings and experimental results conducted with students show that linear regression can achieve up to 80–85% accuracy in predicting final scores, making it an effective tool in educational data mining.