基于BPSO降维的软件故障倾向模块DNN预测

Software fault proneness module DNN prediction based on BPSO dimension reduction

为降低成本并提高软件开发过程的有效性,提出基于束缚态粒子群算法(bound particle swarm optimization,BPSO)降维的软件故障倾向模块深度神经网络(deep neural networks,DNN)预测方法。给出基于BPSO降维的软件故障倾向模块DNN预测算法的计算框架,以及所采用的21个软件故障度量指标和其指标值的归一化预处理方法,采用粒子群算法对软件故障数据集进行降维处理,利用深度神经网络算法实现对软件故障倾向模块的预测。通过在PC1、JM1、KC1和KC3这4组标准测试集上的仿真实验验证了该算法的性能优势。

To reduce the cost and improve the effectiveness of software development process,a software fault proneness module DNN prediction method based on BPSO dimension reduction was proposed.The calculation framework of the software fault proneness module DNN prediction algorithm based on BPSO dimension reduction was presented.The 21 software fault metrics and the normalization preprocessing method of the index value were given.The quantum particle swarm algorithm was used to reduce the dimension of the software fault data set.The deep neural network algorithm was used to predict the software fault proneness module.The performance of the algorithm was verified by simulation experiments on four standard test sets including PC1,JM1,KC1 and KC3.