Integration of neuro-fuzzy modeling in learning management systems to predict academic achievement of graduates

Abstract

This study explores the application of intelligent modeling to support academic decision-making by integrating a predictive system into a university’s learning management environment. Utilizing 25,706 records from 16,158 students over an eightyear period (2015–2022), the dataset includes exam results and final grades across 353 subjects within bachelor's, master's, and PhD programs. After transforming categorical variables - such as education level, course year, and subject name - into numerical format and applying normalization, an Adaptive Neuro-Fuzzy Inference System (ANFIS) was developed to model student performance. This system was chosen for its capacity to capture complex, nonlinear relationships while providing interpretable outputs through fuzzy rules. Comparative evaluation using RMSE, MAE, MSE, and R² metrics demonstrated that ANFIS consistently outperformed alternative models, achieving the lowest RMSE value of 12.80. These findings highlight the model’s reliability and its effectiveness in analyzing academic outcomes across diverse student cohorts. By enabling the early identification of academic risk and delivering interpretable predictions, the system offers practical value to educational institutions aiming to personalize learning pathways and implement data-informed strategies to enhance student success in digital learning environments.