Comparative Analysis of Parameter Estimation Techniques used in Software Reliability Growth Models
Software Reliability Growth Models (SRGMs) have been used by engineers and managers for tracking and managing the reliability of software. The main objective is to achieve the required standard of quality before the software is released to the customer. A number of SRGMs have been proposed in literature to estimate the reliability and quality of software. To apply a given reliability model, defect inflow data is fitted to the selected model equations. All these SRGMs model equations have some unknown coefficients to be estimated. The estimated value of these unknown coefficients can greatly affect the predictions made by these software reliability growth models. There exists two of the widely known and recommended techniques for parameter estimation, maximum likelihood estimation (MLE) and least squared estimation (LSE). Any of these techniques can be used by the software engineers to estimate the values of unknown coefficients. But the estimated values using these approaches differ in most of the cases. So, in this paper a comparative study is performed between these two estimation techniques for their usability and applicability in context of SRGMs. We have also validated the study by predicting the number of failures using different datasets.
Keywords - Software reliability growth models (SRGMs), Maximum Likelihood Estimation (MLE), Least Square Estimation (LSE).