基于动态命令树算法的软件老化趋势预测方法  

Software Aging Trend Prediction Method Based on Dynamic Command Tree Algorithm

作　　者：陈晓璠[1] 邓砚谷[1] CHEN Xiao-fan;DENG Yan-gu(Nanchang Hangkong University,Nanchang Jiangxi 330063,China)

机构地区：[1]南昌航空大学,江西南昌330063

出　　处：《计算机仿真》2021年第11期295-299,313,共6页Computer Simulation

基　　金：江西省高校人文社会科学研究一般项目(GL162034);江西省社会科学“十二五”(2015年)规划项目(15GL41)。

摘　　要：考虑到软件老化具有动态性,准确预测软件老化趋势,可降低软件老化所造成的损失,提出基于动态命令树算法的软件老化趋势预测方法。采用基于降噪自编码器与混合趋势粒子滤波的软件老化趋势预测方法,通过软件未老化特征训练降噪自编码器,把软件老化特征数据输入训练完毕的降噪自编码器中获取重构误差,对重构误差进行平滑处理后,设成观测值输入混合趋势粒子滤波算法,实施软件老化趋势预测。在预测阶段,使用基于动态命令树算法的软件老化趋势预测命令管理方法,实现软件老化趋势预测指令的动态管理,优化软件老化趋势的动态预测效果。实验结果表明,所提方法可提升软件老化趋势预测的精度,预测效果显著。Considering that software aging is dynamic,accurately predicting software aging trend can reduce the loss caused by software aging,a software aging trend prediction method based on dynamic command tree algorithm is proposed in the paper.The software aging trend prediction method based on de-noising self-encoder and hybrid trend particle filter was adopted.According to the non-aging characteristics of the software,the self-encoder was trained to reduce noise.The feature data of software aging was input into the de-noising self-encoder after training to obtain the reconstruction error.After the reconstruction error was smoothed,it was set as the observed value.Then,the hybrid trend particle filter algorithm was input to predict the trend of software aging.During the prediction period,the software aging trend prediction command management method based on dynamic command tree algorithm was used to complete the dynamic management of software aging trend prediction command,and the dynamic prediction effect of software aging trend was also optimized.The experimental results show that the method has high prediction accuracy and excellent prediction effect.

关 键 词：动态命令树 软件老化 趋势预测 降噪自编码器 粒子滤波 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]