神经网络在软件可靠性预测中的应用研究

摘要

Study on software reliability prediction model. RBF neural network is one of the most important software reliability prediction methods, but the search speed of traditional RBF neural network parameters optimization method is low and blindfold, and the software reliability prediction error rate is high, therefore, it is difficult to obtain optimal parameters. In order to improve the prediction accuracy of software reliability, a software reliability prediction model is proposed based on neural network and particle swarm optimization in this paper. Firstly, RBF neural network of initial parameters are used as the particle, and the software reliability and accuracy are used as the objective function of the particle of he particle swarm optimization. Then through collaboration of the particle swarms, the optimal parameters of the RBF neural network are obtained, and finally with optimal parameters of RBF neural network, software reliability is predicted. Using an application database to test and analysis the model, the results show that compared with the traditional RBF neural network, RBF neural network model based on particle swarm optimization has improved the prediction accuracy of the software reliability, the convergence speed is fast, and it is very suitable to software reliability prediction.%关于应用软件保证在应用中安全可靠,应研究软件可靠性预测问题.针对软件可靠性预测系统是一个多因素的、非线性的复杂系统,传统设计高精度的准确数学模型预测方法是相当困难,RBF神经网络是一种非线性预测能力相当强的预测方法.为了提高软件可靠性预测的准确率,提出一种粒子群优化RBF神经网络的软件可靠性预测模型.模型首先将软件可靠性因子作为RBF神经网络的输入,软件可靠性准确率作为RBF神经网络的输出,然后将RBF神经网络的参数初始为粒子群中的粒子,软件可靠性准确率作为粒子优化的目标函数,通过粒子群之间的协作来获得RBF神经网络最优参数,用最优参数对RBF神经网络对软件可靠性进行预测.仿真结果表明,与传统软件可靠性预测方法相比,粒子群优化RBF神经网络对软件可靠性预测的精度更高,收敛速度更快,同时解决了传统RBF神经网络参数寻优难题,更加适合于软件可靠性预测.