Enhancing Software Fault Localization with Variational Autoencoder and Residual Neural Networks

Abstract

Debugging is a critical, costly, and labor-intensive activity in software development. Many fault localization techniques have been proposed to mitigate this issue. Spectrum-based fault localization is a widely used technique that analyzes execution traces (spectra) from test cases and applies a ranking formula to determine the suspiciousness score of each program unit. However, most of the existing spectrum-based fault localization techniques fail to consider complex dependencies between program units and test results. To overcome this limitation, deep-learning-based fault localization techniques have been developed, which utilize artificial neural networks to capture and learn the complex nonlinear relationships between the program spectra and test results. In this study, we propose an effective framework integrating the Variational Autoencoder and Residual Neural Networks (ResNet) to enhance the accuracy of fault localization. First, VAE handles imbalanced input data, and then ResNet networks capture nonlinear relationships in program execution data. Experimental results show that our approach outperforms state-of-the-art techniques in terms of EXAM and RImp metrics.