Neural Network Modeling for Software Reliability Prediction from Failure Time Data
Joong-Yang Park, Sang-Un Lee, Jae-Heung Park · Journal of Electrical Engineering and Information Science · 1999
A variety of statistical software reliability growth models have been developed and utilized. However, since all the statistical models are based on assumptions about the nature of software faults and the stochastic behavior of failure process, there is no single universal model that is applicable for all software projects. Neural networks require only failure data as input and no assumptions are made a priori. This paper considers a neural network approach to software reliability prediction from the failure time data obtained from the software testing. The feedforward network architecture and 4 training regimes are suggested. The cascade-correlation algorithm is used for training and experiments are performed for several data sets collected from different software projects. The experimental results show that feedforward networks adapt well across different softwares and exhibit high accuracy in their predictions.