Enhanced Software Defect Prediction Using Hybrid Stacked Ensemble Learning with XGBoost and Class Balancing

The development of software products to meet precise requirements necessitates adherence to rigorous technical principles and methodologies. Accurate prediction of software defects involves several stages, including data collection, preprocessing, feature extraction, and classification. This study presents an advanced hybrid stacked ensemble model integrating XGBoost with Gaussian Naive Bayes, Bernoulli Naive Bayes, and Random Forest. The ensemble method leverages Principal Component Analysis for feature extraction and incorporates class balancing techniques to enhance predictive accuracy. Extensive Python simulations demonstrate that the proposed model significantly outperforms traditional approaches regarding accuracy, precision, and recall. The results underscore the model's potential to improve software defect prediction and ensure higher software quality. Specifically, the Hybrid Stacked Ensemble model achieved the highest accuracy at 84.6%, outpacing other models such as XGBoost alone, which attained 81.2%. This study is the cornerstone for improved predictive algorithms that accurately identify software faults, enhancing software quality and reliability. Integrating these methods, especially hybrid approaches like XGBoost-SMOTE can improve defect prediction models. Future research should continue to refine these models to address current challenges and optimize their performance in real-world applications.