Enhanced Software Defect Prediction Using a Stacking Ensemble of Decision Tree, Random Forest, and LightGBM

Abstract

To facilitate the discovery of software defects, the latest study used a Stacking Classifier in conjunction with Decision Tree, Random Forest, and LightGBM models. Improved prediction accuracy and stability across a variety of software datasets are achieved by utilizing the complementing properties of many methods. Additionally, the system makes use of a Flask-based front end for safe user authentication and a straightforward user interface for real-time testing and prediction. Software quality assurance becomes much simpler with integration, allowing for faster, safer access, usage, and analysis of faults.