MACHINE LEARNING–BASED BUG DETECTION FRAMEWORK FOR LARGE SOFTWARE PROJECTS

Abstract

Software defects remain a major challenge in large-scale software projects due to increasing system complexity and rapid development cycles. Traditional bug detection techniques often rely on manual code reviews and rulebased testing, which are time-consuming and error-prone. This paper proposes a machine learning–based bug detection framework designed for large software projects. The framework leverages historical defect data, source code metrics, and learning algorithms to automatically identify potential bugs at early stages. Classification and prediction models are integrated into the development workflow. Experimental evaluation is conducted on large software repositories. Results show improved detection accuracy and reduced defect leakage. The proposed framework enhances software quality and development efficiency.