Integrating convolutional neural networks for improved software engineering: A Collaborative and unbalanced data Perspective

Abstract

This study pioneers the tailored application of Convolutional Neural Networks (CNNs) for addressing the challenge of unbalanced data in software engineering, a relatively unexplored domain for CNN utilization. Unlike conventional methods, our framework demonstrates a significant precision uplift of up to 15% in software classification tasks, specifically enhancing minority class sample accuracy. This research not only delineates a novel CNN-based approach that outperforms traditional data balancing techniques but also underscores the strategic integration of AI to bolster software engineering processes. By pinpointing the ethical implications, our findings advocate for a conscientious adoption of AI, ensuring software development advances equitably and efficiently.