组合模型在软件可靠性预测中的建模与仿真

摘要

Software reliability is the results of many uncertainties and multiple linear among the influence factors,so single prediction model cannot fully describe the change rule, which results in the low prediction accuracy of software reliability. In order to improve the prediction precision of software reliability, this paper presents a combination prediction model based on entropy method. First, the model used principal component analysis to eliminate the multiple linear among the software reliability factors and accelerate learning speed. Then AR model and RBF neural network were respectively used to predict the software reliability. Lastly, entropy method was used to determine the weights of two models, thereby the software reliability prediction results were obtained from the combination forecast model. Using NASA's software metric data to test the model prediction performance, the results show that the combination prediction model can improve software reliability prediction accuracy obviously, and is feasible for software reliability prediction.%针对软件可靠性受到多种不确定因素影响,且因素间具有多重共线性,单-预测模型无法全面准确描述其变化规律,导致软件可靠性预测精度不高.为了提高软件可靠性预测的精度,提出一种基于熵值法的软件可靠性组合预测模型.首先采用主成分分析消除软件可靠性度量属性间多重共线性,加快学习速度,然后分别采用AR模型和RBF神经网络对软件可靠性进行预测,采用嫡值法确定两种模型的权重,从而得到组合预测模型的软件可靠性预测值.用NASA的软件度量数据进行模型预测,结果表明,仿真预测模型明显提高了软件可靠性预测精度,说明组合预测方法对软件可靠性预测是可行的.