Research on SRGM Parameter Optimization Based on Improved Particle Swarm Optimization Algorithm

In the field of software reliability, Software Reliability Growth Model (SRGM) is one of the main research methods. Most of the hundreds of models that have been proposed are non-linear function models, and there are certain difficulties in estimating model parameters, and the accuracy of model parameters plays a vital role in the fitting and prediction performance of reliability models. Traditional parameter estimation methods have the problem of easily destroying the constraints on parameter estimation of reliability model and reducing the accuracy of the solution in the optimization process. For this reason, this paper chooses a particle swarm optimization (PSO) algorithm suitable for solving nonlinear optimization problems to solve the model parameter optimization problem. After simplifying the standard PSO algorithm, a new way to construct the fitness function is proposed to estimate the parameters of SRGMs, that is, the function model obtained by making an appropriate mathematical transformation to the maximum likelihood estimation formula of the SRGM parameters. Aiming at five sets of classic software failure data, an improved PSO algorithm is used to solve the parameters of the GO model, and the performance of the model is analyzed through specific software reliability evaluation experiments. The experimental results show that the improved particle swarm algorithm can evaluate the software reliability with high precision and has strong adaptability to the model. Moreover, the evaluation results are significantly better than those obtained by the model using the basic PSO to estimate parameters, and have higher practical application value.