DEVELOPMENT OF AN ADAPTIVE ARTIFICIAL INTELLIGENCE MODEL FOR PREDICTING STUDENTS' ACADEMIC RESULTS BASED ON MACHINE LEARNING METHODS

Abstract

As a result of the digitalization of education systems, the collection of large amounts of academic and behavioral data has expanded the possibilities of predicting students’ academic performance. Early identification of the risk of academic failure or dropout is of great importance in terms of optimizing decision-making mechanisms in higher education institutions. In this research, an adaptive artificial intelligence-based model is proposed for predicting students’ academic outcomes. The model is built on the basis of various features demographic indicators, attendance, current assessment results, and behavioral indicators. In the framework of the study, supervised machine learning algorithms such as Random Forest, Support Vector Machine, and Gradient Boosting were comparatively analyzed and an ensemble approach was applied. Experimental results showed that the proposed adaptive model demonstrates higher prediction accuracy compared to individual models and achieves results higher than 91% on test data. Feature importance analysis showed that attendance and midterm grades collected during the semester are the main predictive factors. The results of the study are of practical importance for the development of early intervention strategies and the optimization of academic management in higher education institutions.