Software reliability prediction based on support vector regression using a hybrid genetic algorithm and simulated annealing algorithm

Software reliability prediction is very important for minimising cost and improving the effectiveness of the software development process. As an important method, relative data during software lifecycle is used to analyse and predict software reliability. However, predicting the variability of software reliability with time is very difficult. Recently, support vector regression (SVR) has been widely applied to solve non-linear predicting problems in many fields such as software reliability prediction and has obtained well performance in many situations, and it is still difficult to select its parameters. Previously, intelligence optimisation algorithms, such as genetic algorithm (GA), are mostly used for finding better parameters of SVR, however existing methods of selecting parameters require usually has some disadvantages. In this study, to overcome weaknesses of GA, such as the local minima and the premature convergence problems, GA and simulated annealing (SA) are integrated into a new optimise algorithm, called GA-SA, it is then applied to SVR for predicting software reliability. The authors compare proposed GA-SA-SVR model with other software reliability models through real software failure data. The experimental results show that the proposed GA-SA-SVR model can obtain better predictions results than the other models and has a fairly accurate prediction capability.