摘要
利用模拟退火算法对动态模糊神经网络的自身参数进行动态调整(SAA-DFNN),并将其应用于软件可靠性增长模型(SRGM)的研究。用软件失效数据在对动态模糊神经网络进行训练的过程中,用模拟退火算法求得动态模糊神经网络自身参数的优化解,根据得到的参数建立基于动态模糊神经网络的软件失效数据预测模型。根据3组软件缺陷数据,将SAA-DFNN建立的SRGM与模糊神经网络(FNN)、BP神经网络(BPN)、G-O模型建立的SRGM的预测能力进行比较,仿真结果表明,根据SAA-DFNN建立的SRGM的单步向前预测能力稳定,预测误差小,并具有一定的通用性。
Simulated Annealing Algorithm(SAA) is used to dynamically adjust the parameters of Dynamic Fuzzy Neural Network(SAA-DFNN).The SAA-DFNN is applied to study Software Reliability Growth Model(SRGM).The SAA is used to resolve the optimal solution of DFNN parameters in the DFNN software failure data training process.Then according to the obtained DFNN optimal parameters SAA sets up software failure data prediction model.Using three groups of software defect data,the predictive ability of the SRGM established by SAA-DFNN is compared with that of the SRGM established by Fuzzy Neural Network(FNN) and by BP Neural Network(BPN)and G-O model.Simulation results confirm that the SRGM established by SAA-DFNN has steady single-step ahead predictive ability with certain versatility and the prediction error is smaller.
出处
《吉林大学学报(工学版)》
EI
CAS
CSCD
北大核心
2012年第5期1225-1230,共6页
Journal of Jilin University:Engineering and Technology Edition
基金
中国科学院知识创新项目(KGCX2-YW-911-2)
关键词
人工智能
软件可靠性增长模型
动态模糊神经网络
模拟退火算法
单步向前预测
artificial intelligence
software reliability
dynamic fuzzy neural network
simulated annealing algorithm
single-step ahead prediction