Towards a Student Dropout Prediction in Higher Education Institutions using a Fuzzy RBF Approach

The serious consequences for people, organisations, and the community at large make the forecast of student dropout in higher education an urgent matter of concern. The issue of student attrition is becoming more complicated as academic systems change, with dropout spanning from failing one course to leaving a whole program. By creating a prediction system that uses preprocessing, feature selection, and model training, our research offers a new way to tackle this difficulty. A GA was used to pick pertinent features, and stringent preprocessing was used to improve data quality. Using fuzzy logic, the system's central component, a Fuzzy RBFNN, efficiently handles the uncertainties and granular patterns common to educational datasets. To further improve the model's classification performance, virtual input variables were added. In terms of accuracy in forecasting student dropout, the experimental findings showed that the suggested strategy earned an average F-score of 98.63%. These results demonstrate how the system can help with student retention by facilitating early detection and intervention strategies. All things considered, a potential way to enhance dropout prediction in higher education and inform targeted policy decisions is to combine fuzzy logic with GA-driven feature selection.