Software Maintainability Assessment Framework Using Object Oriented Design Metrics

Abstract

Software maintainability becomes difficult to evaluate when object-oriented systems contain classes with complexity, coupling, cohesion, inheritance, and modification patterns. This study proposes an Object-Oriented Maintainability Assessment Framework, called OOMAF, to identify classes requiring maintenance attention. The framework organizes information into structural load, dependency pressure, responsibility dispersion, inheritance sensitivity, and change exposure. It normalizes metrics within each project, examines trends across releases, and produces an explainable Maintainability Risk Score with a primary concern for each class. OOMAF was evaluated on ten Java projects containing 3,240 classes across six releases and compared with the Maintainability Index, fixed CK thresholds, and a machine-learning classifier. The proposed framework achieved 87.9% high-maintenance class identification accuracy and an F1-score of 0.86 while reducing false warnings and inspection effort. These findings show that project-relative, temporal, and explainable metric analysis can support refactoring, technical-debt planning, and releasequality monitoring.