摘要
针对线性时间序列方法无法有效预测云工作流活动的运行时间的问题,提出一种基于混沌时间序列的云工作流活动运行时间预测模型。该模型利用相空间重构理论和径向基函数神经网络实现对非线性时间序列的预测。相空间重构理论能够有效刻画云工作流活动的运行时间因受系统性能、网络状况等多种因素影响而呈现的非线性特征;径向基函数神经网络能够有效预测混沌时间序列。模拟实验分别考虑了计算密集型的科学工作流和实例密集型的商务工作流的情况。实验结果表明,无论长周期活动还是短周期活动,混沌时间序列模型明显优于其他有代表性的活动运行时间预测方法。
Aiming at the problem that the linear time series did not efficiently predict the activity durations of cloud workflow, a forecasting model for activity durations in cloud workflow systems based on chaotic time series was pro- posed. The reconstructed phase space theory and Radical Basis Function （RBF） neural network was employed by this model to predict nonlinear time series. The reconstructed phase space theory could depict the nonlinear charac- teristics of cloud workflow due to system performance, network conditions and other factors, and RBF neural net- work was proved to be suitable for predicting chaotic time series. Computation intensive scientific applications and instance intensive business applications were taken into account in simulation scenarios, and the results showed that the proposed chaotic time series model was superior to the existing representative time-series forecasting strategies for both long-duration and short-duration activities.
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2013年第8期1920-1927,共8页
Computer Integrated Manufacturing Systems
基金
国家863计划资助项目(2011AA040501)
国家自然科学基金资助项目(71271071)
中央高校基本科研业务费专项资金资助项目(2012HGBZ0208)
上海高校知识服务平台-可信物联网产学研联合研发中心(筹)资助项目(ZF1213)
武汉大学软件工程国家重点实验室开放基金资助项目(SKLSE2012-09-10)~~
关键词
云工作流系统
混沌时间序列
相空间重构
径向基函数神经网络
时间预测
cloud workflow system
chaotic time series
reconstructed phase space
radical basis function neural net-work
time prediction
