Student success prediction using student exam behaviour

Abstract The Faculty of Mechanical Engineering, Czech Technical University in Prague (FME) faces a significant student drop-out in the first-year bachelor programme, which is an actual problem for many higher education institutions. Metacognitive processes play a vital role in self-regulated learning. Students become active participants in their learning, and one critical aspect of higher education studies is planning and time management. The exam taking behaviour is in the context of the FME manifestation of the time management skills of each student; thus, the exam-taking patterns may help identify at-risk students. To evaluate the importance of exam behaviour patterns, we conducted three experiments. Identification of students passing or failing the first study year has been conducted using four different machine learning models. The exam taking behaviour patterns increase the prediction F-measure significantly for the class of failing students (approximately 0.3 increase). Moreover, the approach based on student behaviour enabled us to identify the critical exam-taking patterns, which further helps the lecturers identify at-risk students and improve their time management skills and chances to pass the first academic year.