In this paper we discuss an experimental evaluation of software reliability analysis using parametric and non-parametric methods. The experimental data set for different, small and large projects were used. We used the normalized root mean square error (NRMSE) as evaluation criteria. The experiments show that the non-parametric models are superior when compared to the parametric models in their ability to provide an accurate estimate when historical data is missing. A comparison among the power, exponential, S-Shape parametric, regression models and the neural network and fuzzy logic models are provided.
展开▼