欢迎来到天天文库
浏览记录
ID:5382316
大小:448.52 KB
页数:7页
时间:2017-12-08
《基于双超平面电力通信网业务qos评价方法》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、第43卷第1期华北电力大学学报Vo1.43.No.12016年1月JournalofNoahChinaElectricPowerUniversityJan.,2016doi:10.3969/j.ISSN.1007—2691.2016.01.16基于双超平面的电力通信网业务QoS评价方法李荣荣,唐良瑞(华北电力大学新能源国家重点实验室,北京102206)摘要:为解决传统业务服务质量(QualityofServices,QoS)评价算法客观性不足、评价效率低等问题。将业务QoS评价映射为业务QoS等级分
2、类,提出了一种基于双超平面(DoubleDecisionHyperplanc,DDH)决策图的电力通信网业务实时QoS评价方法。该方法提取业务网络层特征来表征业务的服务质量,以QoS等级已知的现有业务作为分类器的{IlI练样本,由分类器的构造顺序来确定决策图的形成。此外,算法对传统类间耦合度进行了重新定义,并在决策图各层按照最小类间耦合度原则依次构造分类双超平面,实现业务QoS等级的不完全三分类,避免了传统多分类决策图低层结点“误差崩盘”,自适应性低及决策图结构固定等问题。仿真结果证明,DDH决策图
3、在电力业务QoS评价中比经典DAG-SVM方法具有更短判决时间和更优分类性能,可更好地实现电力业务QoS实时准确评价。关键词:电力通信网;业务服务质量;双超平面;类间耦合度;DAG.SVM中图分类号:TN91文献标识码:A文章编号:1007—2691(2016)01—0092—07QoSEvaluationMethodBasedonDoubleDecisionHyperplanesinElectricPowerCommunicationNetworksLIRongrong,TANGLiangrui(
4、StateKeyLaboratoryofAlternateElectricalPowerSystemwithRenewableEnergySources,NoahChinaElectricPowerUniversity,Beijing102206,China)Abstract:Inordertosolvetheproblemsoftheinsufficientobjectivityandlowevaluationefficiencyintraditionalserv—icesquality(QoS)
5、evaluationalgorithms,anoveldoubledecisionhyperplanes(DDH)decisiongraphmethod,whichmappedtheQoSevaluationtotheQoSlevelclassification,wasproposedtoevaluatethereal-timequalityofservices(QoS)inelectricpowercommunicationnetworks.Itextractedrepresentativefea
6、turestorevealservicesrunningcondi—tionsinthenetworklayerandexploitedtheserviceswhoseQoSwereknowntotraintheclassifiers.Thedecisiongraphstructurewasdeterminedviatheclassgroupingalgorithm,whichformedthegroupsofclassestobeseparatedateachinternalnode.Moreov
7、er,anovelcouplingdegreebetweenclasseswaspresentedandthedoubleplanesineachdecisionlayerwerecreatedbasedonprincipleoftheminimumcouplingdegreebetweenclasses.SincethenodediscriminationsareimplementedviaincompleteternarySVMs,thenewmethodcanwellavoidproblems
8、of‘‘errorcollapse”inlowdeci—sionlayernodes,lowadaptabilityandthefixedstructureinconventionaldecisiongraphalgorithms.Finally,simulationresultsshowthat,comparedwithotherclassicmethods,thenovelmethodexhibitsanumberofattractivemeritssuchase
此文档下载收益归作者所有