EVALUATION AND VALIDATION OF MACHINE LEARNING MODELS TO SUPPORT EDUCATIONAL DECISION-MAKING

Abstract

Artificial intelligence is one of the key factors in the development of modern information systems today. Its integration makes it possible to expand the functionality of traditional information systems, increase the efficiency of data processing and provide support for intelligent decision-making. One of the directions of artificial intelligence development is its integration into the information systems of universities for making managerial decisions. The article presents a comparative analysis of three machine learning models – decision tree, random forest, and gradient boosting – to solve the diagnostic problem of identifying the causes of students' academic failure. In the context of digitalization of education, predictive analytics is shifting the focus from simple forecasting to understanding the factors underlying learning difficulties. The research aims to evaluate models not only in terms of accuracy, but also in terms of interpretability of their results, which is a key factor for their practical application in the decision-making process.

Based on a synthetic dataset simulating the behavior of students in a digital environment, training and model testing were conducted. The results showed high predictive efficiency of all models (accuracy from 98% to 100%). The decision tree has demonstrated 99% accuracy with full transparency of logic, which makes it a valuable tool for generating effective recommendations. XGBoost has achieved 100% accuracy, confirming its status as a stateoftheart algorithm for tasks with clearly structured data.

The paper concludes that it is necessary to choose a model based on the balance between accuracy and interpretability, and outlines the prospects for integrating high-precision models with explicable AI (XAI) methods to create effective decision support systems in education.