Abstract
The article discusses modern approaches to modeling and predicting student academic performance dynamics in educational systems using machine learning methods. The study includes a detailed analysis of existing methodological approaches, ready-made solutions, and commercial platforms; a review of literature sources describing the application of data analysis algorithms in the educational field; and the development of a custom model that covers data collection, processing, model and algorithm selection, and prediction of student performance.Various popular ready-made solutions and commercial platforms that use machine learning methods to analyze, predict, and optimize educational processes have been examined, including: Blackboard Predict, Civitas Learning, Knewton Adaptive Learning Platform, DreamBox Learning, IBM Watson Education, and SAS Campus Analytics. The study analyzed the weights of general attributes that affect prediction, and examined how specific features influence outcomes.The presented article demonstrates that the use of neural networks can significantly improve prediction accuracy, making them a valuable tool for managing educational institutions and making timely administrative decisions. However, one drawback of this algorithm is its long training time on computers with lower computational capabilities. Therefore, other algorithms were also considered during the development of the custom model.The research results showed that ensemble methods significantly reduced prediction errors compared to linear regression, while also requiring much less time for training and forecasting.