Software Quality Assessment Using Explainable Machine Learning

Abstract

Ensuring software quality is critical for the reliability, maintainability, and usability of modern software systems. Traditional software quality assessment techniques often rely on manual reviews, static analysis, or classical machine learning models that offer limited interpretability. This research proposes an Explainable Machine Learning (XML)–based framework to assess software quality by integrating code metrics, defect datasets, and advanced interpretability methods such as SHAP, LIME, and permutation importance. The study evaluates multiple ML models—Random Forest, Gradient Boosting, XGBoost, and Neural Networks—to predict software quality attributes including reliability, maintainability, and defect proneness. Explainability techniques are applied to interpret model decisions, identify key quality indicators, and provide insights useful for developers, testers, and project managers. Experimental results demonstrate that explainable ML improves both predictive performance and decision transparency, making it suitable for practical software engineering environments. This research highlights how combining ML with explainability techniques enhances trust, interpretability, and actionable insights in software quality assessment.