Research On The Predicting Model Of Software Defect Based On Improved BP Algorithm

With the continuous extension of the application of software, the software system security is more and more attached importance to by researchers, and Software testing method also has been more and more attention by engineering personnel. Due to the increase of the testing staff wage costs, the software testing cost problem need to be solved peremptorily. To solve it, the ability of software defect predictions has to be improved.To solve these problems,the research analyze the effect of software defect prediction model based on improves BP algorithm. The article introduces the Research status of software defect model, confirms the software defect model evaluation index, and shows the related research status of BP algorithm. Then the research uses the data of MDP in three different algorithms to do the experiments on software defect prediction model. The three different algorithms include the traditional BP algorithm, the improved BP algorithm with JCUDA, the improved BP algorithm with JCUDA and the simulated annealing algorithm.The research obtained the following conclusions:a. The algorithm of BP can be combined with the software defect prediction model. And it is useful to software defect prediction.b. The algorithm of BP is easy to sink into local minimum result, and costs more time.c. The algorithm combined with JCUDA and BP has the ability to solve the cost of computing time. But it doesn’t solve the problem of local minimum results.d. The algorithm combined with JCUDA, SA and BP solves the problem of local minimum results and improved the accuracy of software defect prediction.This paper shows the related experiments of improved BP algorithm used in software defect prediction model. It proved that JCUDA and SA algorithm has the influence in BP algorithm used in software defect prediction model.