УДОСКОНАЛЕНИЙ МЕТОД ПОБУДОВИ ДЕРЕВ З БАГАТОВИМІРНИМИ ВУЗЛАМИ РІШЕНЬ ДЛЯ ІДЕНТИФІКАЦІЇ СТАНУ КОМП’ЮТЕРНИХ СИСТЕМ

The object of the research is the process of computer system state identification based on the analysis of multidimensional operational data. The subject of the research is methods for constructing decision trees with multidimensional decision nodes and algorithms for optimizing their structure in computer system state classification tasks. The goal of the research is to develop and investigate an improved method for constructing decision trees with multidimensional decision nodes that provides higher accuracy and better generalization ability of computer system state identification models under complex and uncertain conditions. Methods: machine learning techniques, tree-based classification models, nonlinear space partitioning procedures, iterative parameter optimization based on search interval reduction, and classification quality assessment approaches. Results: a procedure for automated selection of the hyperfigure type within a tree node has been developed, enabling the adaptation of the decision boundary to the local structure of data. It has been shown that the use of generalized surfaces reduces the required tree depth and decreases the number of consecutive splits. Experimental studies demonstrate improvements in Accuracy, Precision, Recall, F1-score, and MCC compared with classical decision trees and models with fixed node geometry. Conclusions: the proposed method provides more accurate and stable modeling of boundaries between computer system states while preserving interpretability and maintaining acceptable computational costs. The obtained results confirm the feasibility of applying adaptive multidimensional nodes in monitoring and information security systems.