Abstract
The reliability of information systems is a key prerequisite for the stable operation of modern digital services, especially under high-load conditions and rapidly evolving digitalization processes. Most traditional approaches rely on individual indicators or subjective expert judgments, which limits the transparency of the results obtained. This paper proposes a hybrid methodology that integrates the ISO/IEC 25010:2023 quality model, multicriteria decision-making methods (ARAS, CoCoSo, TOPSIS), and machine learning techniques for imputing missing data. The ISO/IEC 25010:2023 standard is used to define the structure of reliability-related quality attributes and to select appropriate indicators. The AHP method enables a systematic determination of criterion weights based on expert judgments. Missing values in the dataset are restored using a machine learning algorithm, resulting in a complete and consistent data matrix. After data preparation, the reliability of alternative information systems is evaluated using three multicriteria methods. This increases the robustness of the final ranking and reduces its sensitivity to changes in weighting coefficients. The proposed methodology was tested on four real-world information systems with different architectures. A high degree of consistency was observed among the applied methods; the rankings remained stable even when weights were varied by ±10%, and the feasibility of expressing reliability as a percentage was confirmed. This approach enables objective system comparison, identification of weaknesses, and informed decision-making for system improvement and modernization.