Abstract
This study presents a Convolutional Neural Network (CNN) based model for classifying non-defect systems from defect systems to improve software maintenance and reliability. Traditional defect detection methods often depend on manual inspection and handcrafted metrics, which are time-consuming and limited in capturing complex software relationships. To address these challenges, this research employed a data-driven deep learning approach using CNNs to automatically extract hierarchical features from software metrics. The model was trained and evaluated on a publicly available dataset using Python, TensorFlow, and Keras in a GPU-accelerated environment. Results demonstrated exceptional performance, achieving 100% classification accuracy and near-zero loss, confirming the model’s ability to generalize effectively. The proposed approach enables early and precise identification of defect-prone modules, supporting predictive maintenance and reducing operational costs. This work contributes to advancing intelligent, automated software quality assurance and lays the foundation for future research in deep learning-based software defect prediction.

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Copyright (c) 2025 Global Journal of Artificial Intelligence and Technology Development