Computational Assessment of Information System Reliability Using Hybrid MCDM Models

The reliability of information systems (IS) is a key factor in the sustainable operation of modern digital services. However, existing assessment methods remain fragmented and are often limited to individual indicators or expert judgments. This paper proposes a hybrid methodology for a comprehensive assessment of IS reliability based on the integration of the international standard ISO/IEC 25010:2023, multicriteria analysis methods (ARAS, CoCoSo, and TOPSIS), and the XGBoost machine learning algorithm for missing data imputation. The structure of the ISO/IEC 25010 standard is used to formalize reliability criteria and subcriteria, while the AHP method allows for the calculation of their weighting coefficients based on expert assessments. The XGBoost algorithm ensures the correct filling of gaps in the source data, increasing the completeness and reliability of the subsequent assessment. The resulting weighted indicators are aggregated using three MCDM methods, after which an integral reliability indicator is formed as a percentage. The methodology was tested on six real-world information systems with different architectures. The results demonstrated high consistency between the ARAS, CoCoSo, and TOPSIS methods, as well as the stability of the final rating when the criterion weights vary by ±10%. The proposed approach provides a reproducible, transparent, and objective assessment of information system reliability and can be used to identify system bottlenecks, make modernization decisions, and manage the quality of digital infrastructure.