Estimation for Faults Prediction from Component Based Software Design using Feed Forward Neural Networks

Abstract: As far as the software system is concerned, the reliability is one of the important quality attributes in software development process. The recently growing trends in software development process witnessed the paradigm shift to component based software. The reliability of component based software is difficult to estimate directly by taking the reliability of individual components into account and measuring the component reliability in software is not an easy task alike. In this paper we propose a method to estimate the reliability of the software consisting of components by using different neural network architectures. The proposed method considers software consisting of components divided into different sets and observes the number of faults encountered over a cumulative execution time interval for the known set of components and after this we estimate the number of faults predicted for the randomly chosen set of components in software over next cumulative execution time interval. In this process, we estimate the faults prediction behavior in the set of components over a cumulative execution time interval besides this the prediction of faults is estimated for the complete software. We apply the feed forward neural network architectures & its generalization capability to predict the faults in each component of the software with the prediction of faults for the complete software for given cumulative execution time.

Keywords: software reliability estimation, component based software reliability, feed forward neural network, fault prediction.