网络态势预测的广义回归神经网络模型  

GRNN model of network situation forecast

作　　者：卓莹[1,2] 张强 龚正虎[2] 

机构地区：[1]装甲兵技术学院,吉林长春130117 [2]国防科技大学计算机学院,湖南长沙410073 [3]二炮装备研究院,北京100085

出　　处：《解放军理工大学学报（自然科学版）》2012年第2期147-151,共5页Journal of PLA University of Science and Technology(Natural Science Edition)

基　　金：国家973计划资助项目(2009CB320503);国家863计划资助项目(2008AA01A325)

摘　　要：网络态势预测作为网络态势感知的必要环节,能够加强网络管理员对网络状态的认知与理解,为威胁分析和网络规划提供决策支持。在分析现状以及预测方法的基础上,讨论了反向传播、径向基、反馈等神经网络模型用于预测的特点与优势,提出了网络态势预测的广义回归神经网络模型GRNNSF,给出了GRNNSF模型的网络设计原则以及网络态势预测方法。基于真实数据集的实验,验证了GRNNSF模型的准确性和时效性,与其他神经网络模型相比,能更准确地预测网络态势的发展趋势。Network Situation Forecast（NSF）,as a necessary link of network situational awareness,enhances the network cognition and comprehension ability of administrators,and provides the decision support for threat analysis and plan design.Based on the analysis of the research background and forecast method,discussed some Artificial Neural Network（ANN） models and their features and advantages used in NSF,a Generalized Regression Neural Network（GRNN） Model of Network Situation Forecast： GRNNSF was proposed,and the GRNN design principle as well as NSF method presented.An experiment on real datasets was conducted to validate the GRNNSF model.The experiment results show that GRNNSF is more accurate and effective to reflect the network situation trend than other ANN models.

关 键 词：网络态势预测 广义回归神经网络 时间序列分析 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]