Vivi Nur Wijayaningrum, Annisa Puspa Kirana, Ika Kusumaning Putri, Titis Octary Satrio
Vocational Colleges have a higher proportion of practicum courses compared to theoretical because graduates are expected to have the expertise and skills according to their respective fields and are ready to work. Unfortunately, students cannot take short semesters like at university or repeat specific courses in the following semester because of the package system in each semester. Students who get poor grades in practicum courses in one semester have the potential to drop out. However, the number of students who drop out each year will affect the accreditation value of a university. Therefore, this study proposes a system to predict student academic performance in practicum courses based on grades, activity logs, and student backgrounds. A total of 84 data were obtained from the Student Academic Information System (SIAKAD), Learning Management System (LMS), and student surveys. The Multi-Layer Perceptron algorithm is used to form a model that can produce two class outputs, namely students who have the potential to drop out and do not have the potential to drop out. The test results show that modeling using both k-fold cross validation and a training and test data combination gives an average accuracy value and F1-Score above 0.8, which means the classification results are categorized as good. With this good accuracy result, predictions of student academic performance can be used as a decision support tool to prevent dropout students. © 2022 IEEE.
Politeknik Negeri Malang, Department of Information Technology, Malang, Indonesia