基于EMD和GEP的软件可靠性预测模型  

Software Reliability Forecasting Model Based on Empirical Mode Decomposition and Gene Expression Programming

作　　者：张德平[1] 汪帅[1] 周吴杰[2] 

机构地区：[1]南京航空航天大学计算机科学与技术学院,南京210016 [2]东南大学计算机科学与工程学院,南京210096

出　　处：《计算机科学》2013年第4期164-168,184,共6页Computer Science

基　　金：中央高校基本科研业务费专项资金(NS2012072)资助

摘　　要：基于经验模态分解和基因表达式编程算法提出了一种软件可靠性预测模型。通过对软件失效数据序列进行经验模态分解得到不同频段的本征模态分量和剩余分量,消除失效数据中的噪声,运用基因表达式编程算法的灵活表达能力,把分解得到的不同频段的各本征模态分量及剩余分量中所对应的不同失效时间序列作为样本来分别进行预测,重构各本征模态分量和剩余分量中相对应的预测结果,将其作为软件失效的最终预测值。基于两组真实软件失效数据集,将所提出的方法与基于支持向量回归机以及单纯使用基因表达式编程的软件可靠性预测模型进行比较分析。结果表明,该软件可靠性预测模型具有更为显著的模型拟合能力与精确的预测效果。A forecasting method based on empirical mode decomposition（EMD） and gene expression programming（GEP） was presented and applied to software reliability forecasting.Firstly,the software failure samples were handled in order to eliminate the pseudo-data,and the intrinsic mode functions（IMFs） and the residue of different frequency bands were obtained according to EMD.Then the corresponding failure data series in the IMFs and the residue were chosen as the training samples.By means of the flexible expressive capacity of GEP,the models of each IMF and the residue were forecasted.Finally,the ultimate forecasting result was obtained by reconstructing the forecasting results of each IMF and the residue.The method of EMD overcomes the shortcomings that it＇s difficult to select proper wavelet function for wavelet transform,and the final result indicates that the IMFs can reflect the characteristic of software fai-lure.After comparing with the results forecasted by means of combination of SVR and GEP,it proves that the effect of the forecasting method of EMDGEP in software reliability forecasting is better.

关 键 词：经验模态分解 基因表达式编程 软件可靠性预测 可靠性模型 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]