基于稳健回归分析的实时多播协议性能预测方法  

A method for prediction of performance of real-time multicast protocol based on robust regression analysis

作　　者：谭国平 霍孝义 谭林风[1] 倪新洋[1] 刘修泉[2] 

机构地区：[1]河海大学计算机与信息学院通信与信息系统研究所,江苏南京210098 [2]广州番禺职业技术学院,广东广州510000

出　　处：《河海大学学报（自然科学版）》2013年第6期554-559,共6页Journal of Hohai University(Natural Sciences)

基　　金：国家自然科学基金(61001068);广州番禺区科技计划项目(2011-Z-01-07);教育部博士点基金(20100094120017)

摘　　要：为探讨网络编码块长度、节点移动速度、接收节点个数对基于部分网络编码的实时多播(PNCRM)协议性能的影响,提出采用稳健回归分析构建预测PNCRM协议性能回归模型的方法。利用迭代加权最小二乘法求解回归系数,并对该模型进行显著性检验。结果表明,该模型对PNCRM协议性能预测的平均相对误差仅为4.83%,因而具有较准确的预测能力。To investigate the influences of the BlockSize,the speed of node movement,and the number of receivers on the performance of partial network coding-based real-time multicast protocol（ PNCRM）,a model for the prediction of the performance of PNCRM was built based on robust regression analysis. The iteratively reweighted least-squares algorithm was used to solve regression coefficients. A significance test was conducted for the model.The results show that the mean relative error was only 4. 83% when using the model for the prediction of the performance of PNCRM,and,thus,the model has strong prediction capability.

关 键 词：稳健回归分析 迭代加权最小二乘法 预测模型 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]