Measures for Predicting Software Reliability using Time Recurrent Neural Networks with Back-propagation
Manmath Kumar Bhuyan, Durga Prasad Mohapatra, Srinivas Sethi · ACM SIGSOFT Software Engineering Notes · 2015
Recurrent Neural Network (RNN) has been known to be very useful in predicting software reliability. A number of parametric models and reliability growth models, have been proposed, but developing a model that can predict reliability in all types of data sets, in any environment, and at any phase of software development is still a challenge. In this paper, we propose a model that explores the applicability of Recurrent Neural Network with Back- propagation Through Time (RNNBPTT) learning rule to predict software reliability. The detailed procedure of reliability prediction using recurrent neural networks is explained. The model has been applied on data sets collected across several standard software projects during system testing phase with fault removal. Though the procedure is relatively complicated, the results depicted in this work suggest that Fully Recurrent Neural Networks (FRNN) exhibits an accurate and consistent behavior in reliability prediction.