Multi-Layered machine learning model for mining learners academic performance

Abstract

Different colleges and universities have different approaches to dealing with low-performance learners. However, in most cases, analgesics do not deal with root problems. This research suggests a model of three layers of variables sequentially adaptable to a deep-root issue. The suggested model can identify early pupils who could be at risk because of inaccurate or lack of match sequences and suggest rehabilitation. The approach proposed was implemented at three levels. First, we examined the personality type for 180 learners from different majors: Security and Forensics, Networking, and Application Development, using the MBTI test. Second, we build a knowledge matrix for courses by dividing each learning outcome into its knowledge segments. Then, we build the skills matrix for courses by decomposing each learning outcome into its skills segments. We then use machine learning (SVM, DT and association rules) algorithms to mine student performance on a smaller scale of knowledge and skills, taking into account their personality types instead of measuring an entire course's holistic performance. Finally, we developed a system of recommendations to detect performance deviations in knowledge and skills and provide adaptive learning materials that fit the examined students' personality. The proposed approach demonstrates its validity and effectiveness. However, it needs regular updates on learners' performance, which could be automated and linked to evaluation tools. The framework also has a minor impact on learners' privacy since it exposes individual personalities to their advisors.