Early Detection of Learning Abnormalities in Higher Education Students: A Hybrid Decision-Making Approach

Recently, the pandemic has caused tremendous changes in teaching learning and career attainment gaps in higher institution students. Earlier approaches to student performance predictions do not deliver the expected quality in prediction performance, given the enormous quantity of student data accessible to educational institutions. The performance literature has examined machine learning (ML) strategies to enhance predicting skill. This research proposed hybrid machine learning approach (FEPRP) that integrate unsupervised (K-medoids) and six conventional supervised machine learning algorithms separately to predict student academic performance. We use several methodologies and compare the hybrid model's performance to that of separate supervised ML models. In terms of forecasting student achievement, hybrid models beat their solo counterparts. Furthermore, whereas the existing research overlooks prior performance, we discover that adding these features enhances the expected efficiency of the suggested mixed model (FEPRP).