An analysis of code quality metrics applying correlation and regression for identifying bug occurrences
Year: 2025
Type: article
Abstract: This study explores the relationship between software quality metrics and defect proneness in the Eclipse project using regression analysis. The results reveal that key metrics such as Lines of Code (LOC), Weighted Methods per Class (WMC), and Response for a Class (RFC) significantly impact bug occurrence, indicating that larger and more complex classes are more prone to defects. Additionally, lack of cohesion (LCOM) is found to be a strong predictor, reinforcing the importance of well-structured class design. The regression model explains approximately 40% of the variance in bug occurrences, highlighting its predictive power. Findings emphasize that high-cohesion, low-coupling principles should guide software development to enhance maintainability and reliability. Prioritizing classes with high-risk metrics for rigorous testing and refactoring can significantly improve software quality. Future work will extend this analysis to other projects and refine predictive models to enhance defect detection strategies. The paper is structured as follows: it begins with an introduction to software quality, followed by a review of related work. The third section outlines the adopted methodology, while the fourth presents the metrics. Finally, the fifth section discusses key findings and offers recommendations for improving software quality.
Authors: Madhumita Singha, Mukesh Kumar Shrivastava, Vandana Bhattacharjee
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4409471817