Effective Software Fault Localization using an Back Propagation Neural Network

Abstract

Effective fault localization is an essential requirement of software development process. Backpropagation neural network can be used for localizing the faults effectively and efficiently. Existing NN based fault localization techniques take statement invocation information in binary terms to train the network. In this paper, we have proposed an efficient approach for fault localization using back propagation neural network and we have used the actual number of times the statement is executed to train the network. We have investigated our approach on Siemens suite. Results show that on an average there is 35% increase in the effectiveness over existing BPNN.