Research On Parameter Estimation Of Software Reliability Model Based On PSO And ABC Hybrid Algorithms

The vigorous development of computer science and technology makes our lives can not be separated from a variety of complex software,the quality of these software must be guaranteed.Software reliability is the main quantitative standard to evaluate the quality of software and it can be predicted by software reliability model.However,it is difficult to estimate the parameters of the model due to the nonlinearity of the model.A new idea is to use the group intelligence optimization algorithm for parameter estimation of software reliability model,and the key to the realization is to construct the fitness function.Therefore,how to construct a reasonable and effective fitness function is very necessary.In view of the shortcomings of the traditional fitness function,such as large range of initialization,slow convergence speed and low accuracy.In this paper,a new construction method of fitness function is proposed to estimate the parameters of software reliability model,that is,to make a proper mathematical transformation for the maximum likelihood estimation formula of the software reliability model parameters.Based on the new fitness function,the accuracy of the parameter estimation can be further improved by eliminating the obvious inappropriate solution and adding prior knowledge in the implementation of the algorithm.This paper has discussed the basic theroy of software reliability,software reliability model,PSO algorithm and ABC algorithm respectively from the perspective of theoretical analysis and compares the advantages and disadvantages of two algorithms in the estimation of the parameters.Particle swarm algorithm is simple to implement and has fast convergence speed,but the accuracy of the solution is not high because it is easy to premature convergence in the latter of the iteration;Artificial colony algorithm has strong searching ability and high accuracy,but the algorithm is more complex and will be at the expense of the amount of calculation.In view of the advantages and disadvantages of PSO algorithm and ABC algorithm,this paper proposed a hybrid algorithm PSO-ABC to estimate the parameters of software reliability model.The basic idea is to introduce the artificial colony search operator into the particle swarm algorithm.After the particle swarm algorithm has completed the basic search,the artificial colony search operator is used for further searching around,which not only preserves the simplicity and realizability of the PSO algorithm,but also enhance its ability to explore.At last,based on the new fitness function and classical software failure data sets,this paper uses PSO algorithm,ABC algorithm and their hybrid algorithm respectively to estimate the parameters of GO model,and then make the prediction and comparison.The experimental results show that compared with the old fitness function,the new fitness function is simple and easy to implement,it can more accurately and efficiently to estimate the parameters of the software reliability model,and has better software reliability prediction effect.Compared with a single algorithm,the hybrid algorithm has higher accuracy and is not at the expense of the computational cost,it has achieved better results no matter in parameter estimation or model prediction.