Intelligent models and sustainability assessment of the security system of agro-industrial enterprises

Background. In the era of digitalization, maintaining resilient security systems in agro-industrial enterprises is crucial. This paper examines approaches to developing intelligent models aimed at assessing and predicting the resilience of organizational and technical systems based on the analysis of interrelated risk factors. Cognitive and fuzzy modeling approaches are applied as methodological tools to formalize expert knowledge and support managerial decision-making. A methodology for constructing an integrated resilience indicator that takes into account both external and internal dynamics is proposed. Scenario analysis demonstrates the potential of intelligent algorithms to model critical situations and to select optimal response measures. The developed models can be applied to strengthen infrastructure protection strategies, enhance information and physical security, and ensure the sustainable operation of enterprises in uncertain environments.

The aim of the study is to develop and verify a model based on fuzzy cognitive maps (FCMs) for the mathematical assessment of the resilience of agricultural enterprise security systems. The work aims to integrate expert knowledge, scenario modeling, and dynamic visualization of system behavior under changing external and internal factors.

Materials and methods. The methodological framework of the study is based on cognitive and fuzzy modeling, simulation, and machine learning. FCMs are used as tools, accounting for uncertainty, the subjectivity of expert assessments, and nonlinear relationships between factors. Logistic Regression, Random Forest, and XGBoost algorithms, implemented in Python, were used for computational experiments. The analysis was conducted using the IGLA package for constructing cognitive models and assessing impact scenarios.

Results. An intelligent security system resilience model was developed, incorporating five key concepts: financial resilience, human resources, technological reliability, information security, and organizational processes. Scenario modeling was conducted to identify the impact of various management strategies on the integrated resilience indicator. Scenario simulations revealed that an integrated approach can increase overall system resilience by 15–20% compared to isolated security improvements.

Machine learning experiments achieved a high classification accuracy (up to 0.98) across all models, with logistic regression providing the best balance between precision and recall.

Conclusion. Intelligent models based on fuzzy cognitive maps and machine learning methods provide effective assessments of the resilience of security systems in agricultural enterprises. The proposed approach allows for the consideration of uncertainty, modeling threat scenarios, and improving the adaptability of security systems. The practical significance of this work lies in the potential application of the developed models to improve infrastructure protection strategies, enhance information and physical security, and ensure the stable operation of enterprises in uncertain environments.