An integrated approach to forecasting software system quality using bayesian correction, multi-criteria optimization, and meta-learning
Year: 2025
Type: article
Abstract: The aim of this study is to improve modern methods for forecasting the quantitative characteristics of software system quality to enhance reliability, efficiency, and adaptability in dynamic IT environments. To achieve this, an integrated forecasting approach was developed that combines adaptive Bayesian noise correction, probabilistic ensembles with weighted risk adjustment, hybrid multi-criteria optimization, graph models of metric interdependencies, and meta-learning for forecast adaptation. The scientific novelty lies in the proposed ensemble integration and adaptive mechanisms that increase forecasting robustness while accounting for uncertainty and metric dependencies. The methods were validated on the GitLab CE system. The experimental results confirmed measurable improvements: forecasting error was reduced from 18.7% to 4.2%, execution time decreased by 36.8%, CPU and memory consumption dropped by up to 20%, and system reliability indicators (downtime, fault tolerance) improved by more than 60%. These metrics confirm that the proposed approach strengthens reliability, efficiency, and adaptability of software quality forecasting compared to traditional methods.
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $107
OpenAlex ID: W4415547360