Approach to Intelligent Support in Educational Program Design Using X-Matrices and a Domain-Specific Language

This paper presents an automation system for the design and evaluation of educational programs using the X-matrix approach and natural language processing technologies. A processing cycle was developed to provide recommendations. The evaluation of machine learning classifiers identified Random Forest, XGBoost, Naive Bayes, Light GBM, and CatBoost as optimal models. Features derived from similarity metrics between text descriptions were assessed, with the most suitable feature set selected. The achieved metrics demonstrated satisfactory precision, recall, and F1-score values. The developed models were utilized to predict elements in cross-matrices for new educational programs, providing logical explanations for the course-competency connections. Methodologists can make informed decisions based on the model's arguments. This research contributes to the automation of educational program design and offers valuable insights for improving program coherence and technology integration.