Tracking and forecasting of students progress using Multiple-Criteria Decision Making and kNN methods

Abstract

The paper discusses a problem in higher education related to students' progress and performance in exams. During the semester, every student would benefit from information about whether their work on the course is sufficient to achieve the desired result on the final exam. If it was pointed out to them at the right moment that success in the subject is likely to be absent, it could encourage them to increase efforts for the rest of the semester and thus in the end achieve their goal. However, if a bad result is expected for most participants, a change in the teaching method should be considered. Therefore, in this research, two aims have been set: to determine the moment during the semester when it can be said that a certain student will not successfully complete the course and to check that the teaching methods are appropriate. In order to give advice to the student in a timely manner, at certain moments during the semester, it is necessary first to diagnose the level of their knowledge, and then make a prediction of the expected result on the exam. In the solution proposed in this paper, in five time moments (t2,...,t6) the student's knowledge is assessed using the Multiple-Criteria Decision Making (MCDM) method - PROMETHEE II, which considers four criteria (attendance, activity, homework, and test), and the teacher's preferences. Thus, a progress function for a given student is obtained, which shows how they are progressing toward the set goal. Based on the values of the progress function in milestones t2-t6, student achievement is categorized through grades. In this investigation, we have used the k-Nearest Neighbors (kNN) Algorithm for grade classification. Our findings clearly justify that MCDM and kNN models are appropriate for tracking, classifying, and forecasting student grades.