Authors: Assistant Professor G.Preethi, K . Arundhathi
Abstract: The rapid growth of software applications across industrial, commercial, and scientific domains has significantly increased the demand for reliable and high-quality software systems. , identifying defects during the early stages of development has become an essential requirement for reducing maintenance costs, improving software reliability, and enhancing overall product quality. Traditional software testing approaches primarily depend on manual inspection and extensive debugging procedures, which are often time-consuming, resource-intensive, and incapable of detecting all potential defects before software deployment. Recent advancements in Machine Learning (ML) have introduced intelligent data-driven techniques capable of analyzing historical software metrics to predict defect-prone modules, thereby supporting proactive quality assurance throughout the software development life cycle. This study presents an intelligent proposed framework utilizes software engineering metrics collected from the JM1 repository to train multiple supervised machine learning models for identifying defective software modules. The framework incorporates comprehensive data preprocessing, including label encoding, one-hot encoding, feature scaling, and dataset standardization before model training. Seven supervised learning algorithms, enabling objective evaluation of each classifier's predictive capability. The proposed framework is evaluated using widely accepted performance measures including Accuracy, Precision, Recall, F1-Score, and Root Mean Square Error (RMSE). Experimental analysis demonstrates that the Random Forest classifier provides the highest overall prediction performance, achieving approximately 81% classification accuracy together with superior precision, recall, F1-score, and the lowest RMSE among the evaluated models. methodology for software defect prediction. The developed framework offers a reliable, reducing maintenance effort, and supporting more efficient software engineering practices.
