Estimation of reliability parameters of software growth models using a variation of Particle Swarm Optimization

Reliability of any software product is a quantifiable attribute which is essential for predicting the degree of credibility of the software to operate accurately for a specific period of time without the occurrence of any kind of failure. Prediction of the behaviour of the software before its final shipment is an important task and behaviour includes satisfactory performance which is largely depends on reliability of the software. Various software reliability growth models have been proposed to assess the reliability of the software. An optimized estimation of parameters of software reliability growth models is the matter of concern as the accurate prediction of reliability depends on these parameters. Although the traditional methods like MLE and LSE are capable of evaluating these parameters but normally these parameters possess nonlinear relationships which become problematic in finding the optimal parameters to tune the model for a better prediction. A Swarm Intelligence Based stochastic search techniques named as Particle Swarm Optimization has been adopted in this work for the evaluation of growth models which presents better and optimized results also it helps avoiding problems that used to occur while estimating software reliability growth parameters using traditional methods. Particle Swarm Optimization will be used along with some modifications for estimation of the NHPP based Reliability Growth Models.