网络运行态势感知技术及其模型  

Network running situation awareness technology and its model

作　　者：柏骏[1] 夏靖波[1] 钟赟[1] 陈珍[1] 

机构地区：[1]空军工程大学信息与导航学院,陕西西安710077

出　　处：《解放军理工大学学报（自然科学版）》2015年第1期16-22,共7页Journal of PLA University of Science and Technology(Natural Science Edition)

基　　金：陕西省自然科学基础研究计划重点资助项目(2012JZ8005)

摘　　要：针对现有网络态势感知技术无法反映网络整体运行质量的问题,提出了网络运行态势的概念,并给出其模型。首先,将马尔科夫链的转移概率引入网络链路权重评价中以衡量相邻节点的可达性,进而求得网络链路权重;随后,为一致、稳定地评估网络运行质量并克服基于支持向量机评估方法在训练过程中存在的参数难以确定、过拟合等问题,提出了基于相关向量机的网络运行质量评估方法;最后,将各链路的运行质量评估值按照权值加权得到网络运行态势,实现网络运行质量局部评估向网络运行态势整体呈现的提升。实验结果表明,该模型获得的网络态势值与通过灰色模糊综合评估得到的态势值的平均绝对误差仅为1.61%,优于基于支持向量机的网络评估方法;通过对态势均值及方差的比较,可对不同网络建设方案提供必要的技术指导。The current network situation awareness technologies are not able to reflect the whole network running quality,so the concept of network running situation was proposed and the network running situation awareness（NRSA）model established.Firstly,the Markov chain model was introduced to the weights evaluation of network links to measure the reachability of adjacent nodes.Secondly,for evaluating network running quality consistently and smoothly,a network running quality evaluation method based on relevant vector machine（RVM）was proposed,thus overcoming some problems of the method based on support vector machine（SVM）.Finally,network running situation was acquired through the weighted sums of running quality evaluation value of every link,realizing the promotion from partial link running quality to holistic network running situation.Experimental studies illustrate that the network situation acquired by the proposed model has a mean absolute percent error（MAPE）of 1.61% compared with that acquired by grey fuzzy comprehensive evaluation,which is better than that of the network evaluation method based on SVM.A guidance can be presented by comparing the situation average and variance of different network construction.

关 键 词：网络态势感知 运行态势感知 链路权重 运行质量评估 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]