МЕТОДЫ ИНТЕРПРЕТАЦИИ ЧЕРНЫХ ЯЩИКОВ МАШИННОГО ОБУЧЕНИЯ И ИХ ПРИМЕНЕНИЕ ДЛЯ СОЗДАНИЯ СИСТЕМ ПОДДЕРЖКИ ПРИНЯТИЯ РЕШЕНИЙ

Abstract

Artificial intelligence (AI) has made great strides recently with the rapid development of machine learning technologies. Despite this, there are potential risks associated with a black-box approach to learning. Unlike some classical machine learning methods, especially decision trees, in which model results can usually be explained, nonlinear classification and especially deep learning models lack transparency, making it difficult to understand how a model made a particular decision. Recent models for explaining machine learning results partially solve this problem. The goal of the research community is to create explicable machine learning models (EMLs). An important outcome of this effort is the possibility of applying EML to decision support systems. This paper describes some state-of-the-art models for explaining the results of machine learning models and describes an approach for applying EML to multicriteria decision support systems. The essence of the approach described is to build a machine learning model based on data about control objects and then use the explanations to make recommendations for influencing the objects to improve the target. We present a scheme of such a system and examples of the results of its application in the management tasks of secondary education and health care organizations.

More by Yan Kuchin

• MAJOR SOURCES OF FINE PARTICULATE AIR POLLUTION IN FOOTHILL MEGACITY: THE CASE OF ALMATYHydrometeorology and ecology · 2026
• EXPLAINABLE MULTISPECTRAL ATTENTION NETWORK FOR SENTINEL-2 AGRICULTURAL LAND CLASSIFICATIONScientific Journal of Astana IT University · 2026
• SEMANTIC SEGMENTATION OF AGRICULTURAL FIELD ANOMALY PATTERNS IN AERIAL IMAGES USING U-NETHerald of Kazakh-British technical university · 2026
• UAV-Based Classification of Crop Phenological Stages Using Deep LearningDrones · 2026