<![CDATA[Software defect classification is a crucial process in quality assurance, pivotal for the development of reliable software systems. This paper presents an innovative approach that synergizes traditional software complexity metrics with advanced machine learning algorithms, namely Light Gradient Boosting Machine (LightGBM) and Extreme Gradient Boosting (XGBoost), to enhance the accuracy and efficiency of software defect classification. Leveraging a dataset characterized by McCabe's and Halstead's metrics, this study embarks on meticulous data preprocessing, feature engineering, and hyperparameter optimization to train and evaluate the proposed models. The LightGBM and XGBoost models are fine-tuned through the Optuna framework, aiming to maximize the ROC-AUC score as a measure of classification performance. The results indicate that both models perform robustly, with XGBoost demonstrating a slight superiority in predictive capability. The integration of machine learning with traditional complexity metrics not only enhances the defect classification process but also provides deeper insights into the factors influencing software quality. The findings suggest that such hybrid approaches can significantly contribute to the predictive analytics tools available to software engineers and quality assurance professionals. This research contributes to the field by offering a comprehensive methodological framework and empirical evidence for the effectiveness of combining machine learning algorithms with traditional software complexity metrics in software defect classification.]]>
Copyrights © 2024
