Predicting software reliability growth using nonparametric regression

In this paper, we use two Nonparametric Regression (NPR) methods to predict the growth of software reliability. These are the Nadaraya-Watson (N-W) and the Local Linear (LL) Estimator. The main advantage of using these methods is that they place minimum requirement on the distributional form of the stochastic process which gave rise to software failure data and hence dispense with the need to estimate parameters from complex models. Sample size consideration based on using the two NPR methods will also be considered in this paper. Finally, numerical examples involving four sets of real software data v.-ill be presented to illustrate the techniques presented in this paper.