Exploring Simulated Annealing Algorithm for Parameter Estimation of Software Reliability Models

Software reliability is significant to the aircraft airworthiness. Software reliability growth models (SRGMs) are very important for software reliability estimation and prediction and have been successfully applied in software reliability engineering. The parameters of SRGMs can characterize the model's behavior and have a physical interpretation. Hence, the approaches used to the parameter estimation also play an important role in software reliability modeling and then should have a further discussion. Simulated Annealing (SA) algorithm has been popularly used to solve various optimization problems. In this paper, a modified SA algorithm for parameter estimation (MSAE) of SRGMS is presented. Two case studies are then presented and analyzed for comparing the fitting and the convergence performance of MSAE with LSE on several SRGMs and several failure data-sets. The results show that compared with LSE, SRGMs with the MSAE has a comparable at least or a little better fitting performance and a notable better convergence performance for parameter estimation.