关于软件可靠性预测仿真的研究方法

A Research Method of Software Reliability Prediction and Simulation

软件可靠性预测已成为当前软件工程的一个研究热点。现有的多模型变权重组合的软件可靠性预测模型由于易忽视故障排除率函数和故障引入率函数的影响,且各模型权值不能精确反映在组合模型中的作用,无法对权值进行动态调整,使得软件可靠性预测精度不高。针对上述问题,提出了采用多模型变权重组合的软件可靠性预测新方法。建模过程中,考虑到故障排除率函数和故障引入率函数对软件可靠性预测精度的影响。且采用熵值法确定组合模型中能精确反映各模型作用的权值,依据各模型作用效果对权值大小进行动态调整,进一步提升了预测精度。最后,引入了一组公开发表的软件失效数据,验证新组合模型的预测效果。通过与已有组合模型对比,说明新组合模型具有更好的预测精度和拟合能力。

A new software reliability prediction method for combination of multi - model based on changeable weight based on a variety of software reliability prediction models is proposed. During the established model, the influence of fault removal rate function and fault introduction rate function are taken to software reliability prediction accuracy into account. And each model＇s weight which can accurately reflect the effect on the combination of multi - model is determined by the entropy method, it could be dynamically adjusted according to the effect of each model＇s function, and the prediction accuracy is further improved. At last, a set of publicly published fake data is introduced to test the predictive effect on the new combination of multi - model. The result shows that the new model has better prediction accuracy and fitting ability than the existing mixed model.