Software Maintainability Index Prediction with Source Code Metrics using Machine Learning
Kökten Ulaş Birant · DergiPark (Istanbul University) · 2026
Software Maintainability Prediction (SMP) plays a significant role in software engineering, as it provides developers and project managers with insights into quality, potential maintenance costs, and scalability over time. In this research, a machine learning–based model for SMP is presented using the M5P Model Tree algorithm. The prediction is performed based on source code metrics, including comment density, cyclomatic complexity, Halstead volume, and cohesion. Experiments conducted on a dataset demonstrated that the trained model achieved a Correlation Coefficient (R) of 0.8892, indicating a high predictive capability. Compared to its counterpart methods, it delivered a 9.23% improvement in terms of R, confirming its superior performance.