摘要
路由节点行为预测可以为网络安全管理以及路由行为评估机制等提供重要的决策依据,而现有的路由节点行为预测算法存在诸如依赖专家经验、对样本要求较高以及在节点行为波动较大的情况下预测准确度下降等问题.为克服上述问题,该文在灰色预测模型的基础上提出了一种路由节点行为预测算法,该算法将路由节点行为序列中的波动类型被分为突发波动和迁移波动,结合Markov预测模型实现波动类型识别,随后基于不同的波动类型设立相应的未来行为值预测方法:对于突发波动,预测方法基于平滑级比序列的灰色预测结果;对于迁移波动,预测方法基于路由节点行为序列的灰色预测结果.最后,使用Markov模型对预测结果进行修正.实验结果表明,相比于已有的节点行为预测算法,该文的预测算法在预测精度上有较大提升.
The routing node behavior prediction can offer important decision basises for network security management and routing behavior evaluation mechanism, but the current prediction algorithms have faced some problems such as dependence on expert experience, high sample require ment and low prediction accuracy for the node behavior with large fluctuation. In order to over come the above problems, this paper presents a prediction algorithm based on Grey prediction. In this algorithm, the fluctuations in the routing node behavior series are classified as burst fluctua- tion and shifting fluctuation, and the fluctuation type identification is calculated by using Markov model. Then corresponding prediction methods for future behavior value are setted based on dif- ferent fluctuation types： as to the burst fluctuation, the prediciton method is based on the grey prediction result of the smooth ratio series; as to the shifting fluctuation, the prediciton method is based on the grey prediction result of the routing node behavior series. At last, Markov model is used to correct the error of the result. The experimental result shows that the prediciton accuracy of our prediction algorithm is higher than the other routing node behavior prediction algorithms.
出处
《计算机学报》
EI
CSCD
北大核心
2014年第2期326-334,共9页
Chinese Journal of Computers
基金
国家"九七三"重点基础研究发展规划项目基金(2010CB328104)
国家自然科学基金(61003257
61070210
61272531)
国家"八六三"高技术研究发展计划项目基金(2013AA013503)
国家科技支撑计划课题(2010BAI88B03
2011BAK21B02)
高等学校博士点专项科研基金(20110092130002)
江苏省网络与信息安全重点实验室资助项目(BM2003201)
教育部计算机网络与信息集成重点实验室(东南大学)资助项目(93K-9)资助~~
关键词
路由节点行为预测
灰色预测模型
波动类型识别
网络行为
routing node behavior prediction
grey prediction model
fluctuation type identifica tion
network behavior
