The use of generative artificial intelligence (AI) to analyze educational data and predict student academic performance

Abstract

The relevance of the topic of using generative artificial intelligence (AI) to analyze educational data and predict student academic performance is due to the growing digitalization of higher education and the need for new tools to improve its effectiveness. An analysis of modern publications has revealed a lack of research into the possibilities of generative AI in this subject area. The aim of the work is to develop and test a model based on generative adversarial networks (GAN) for analyzing arrays of educational data and predicting individual trajectories of academic success. Tasks: 1) adapt the GAN architecture to the specifics of educational data; 2) train the model on a representative sample from the LMS logs of 5,000 students; 3) verify predictive accuracy on independent data; 4) compare the results with traditional machine learning methods. Methods: generative adversarial modeling, Bayesian optimization of hyperparameters, stratified k-block cross-validation, Friedman and Nemeni tests. Results: 1) the constructed GAN model surpasses the basic solutions in accuracy and completeness; 2) is resistant to class imbalance and retraining; 3) allows you to generate synthetic profiles of students to expand the set of features; 4) demonstrates practical applicability in the real educational process. The proposed approach opens up prospects for the creation of adaptive learning systems capable of identifying students at risk in real time and providing individualized support.