Software Defect Prediction Using Temporal Transformer Graph Network with Newton–Raphson Optimization

Rajesh Kumar Udumu; D. Vasumathi

doi:10.37965/jait.2025.0751

Software Defect Prediction Using Temporal Transformer Graph Network with Newton–Raphson Optimization

Authors

Rajesh Kumar Udumu Department of Computer Science and Engineering, JNTUH https://orcid.org/0009-0009-3161-7816
D. Vasumathi Department of Computer Science and Engineering, JNTUH

DOI:

https://doi.org/10.37965/jait.2025.0751

Keywords:

data balancing, graph convolutional networks, Newton–Raphson optimization, risk management, software defect prediction, temporal transformer attention

Abstract

Software Defect Prediction (SDP) is crucial for ensuring software reliability by identifying fault-prone components early in development. Traditional statistical and machine learning models, though effective to some extent, struggle with data imbalance, noise sensitivity, and limited feature representation. Recent advances in deep learning, including CNNs, RNNs, and GCNs, improved learning capabilities but still face overfitting and high computational costs. To overcome these challenges, this research introduces a Temporal Transformer Graph Convolutional Network with Newton based Raphson Optimization (TTRGCN-NRO) deep learning framework that effectively captures spatial-temporal dependencies and enhances prediction accuracy, convergence speed, and interpretability in software defect prediction.