软件项目风险评估算法的仿真研究

Application of RBF Neural Network in Risk Assessment of Software Project

研究软件项目风险评估准确性问题,软件项目具有复杂性,软件风险因子具有不确定性和非线性,采用传统数学方法无法建立准确软件项目风险评估模型,同时由于受到人为的主观因素的影响,导致软件项目风险评估准确率比较低。为了提高软件项目风险评估准确率,提出一种RBF神经网络的软件项目风险评估模型。模型首先采用专家系统构建软件项目风险评估的指标集,再用德尔菲法对评估指标打分,最后将指标得分和风险等级分别作为RBF神经网络输入和输出,通RBF神经网络的智能学习得到最优软件项目风险评估模型。在MATLAB平台对软件项目评估数据集进行仿真,结果表明,RBF神经网络模型克服了传统数学评估模型的缺陷,提高了软件项目风险评估的准确率,降低软件项目风险评估误差,提供了一种软件项目风险评估的依据。

Research on software project risk assessment problem.Because the software project has uncertainty factor,the traditional models cannot accurately set up software project risk assessment model,resulting in software project evaluation large error.In order to reduce software project risk evaluation error,a software project risk assessment model is proposed based on RBF neural network.The model constructs a software project risk assessment index collection which is divided into several statures,uses Delphi method to evaluate index score,and then is trained and modeled by the RBF neural network to get the best software project risk assessment model.Finally,in MATLAB platform,the model is tested by the software project risk assessment examples.The simulation results show that the model improves the accuracy of software project risk assessment and reduces software project risks.