基于聚类与贝叶斯分类器的网络节点分组算法及评价模型  

Internet Hosts Grouping Algorithm Based on Clustering and Bayesian Classifier and Evaluation Model

作　　者：翟健宏[1] 李伟[1] 葛瑞海[1] 杨茹[2] 

机构地区：[1]哈尔滨工业大学计算机网络与信息安全研究中心,哈尔滨150001 [2]黑龙江工程学院计算机科学与技术系,哈尔滨150001

出　　处：《电信科学》2013年第2期51-57,共7页Telecommunications Science

基　　金：国家自然科学基金资助项目(No.61173145);国家重点基础研究发展计划("973"计划)基金资助项目(No.2011CB302605)

摘　　要：多组群网络应用是目前比较流行的互联网服务模式之一,本文主要针对多组群网络应用的节点分组及评价方法进行研究探讨,设计了一种基于无监督聚类方法和贝叶斯分类器的高质量分组算法Clustyes,提出了基于"组耦合度"的可调多取向分组评价模型,该模型较好地将组内节点间往返时延、分组丢失率和跳数3个指标对组内通信的影响有机地量化融合。Clustyes算法在NS-3上进行了模拟,结果表明,使用该算法获得的分组具有组内节点通信性能好、稳定性高、物理距离近的优点。The multi-group network application is one of the popular network service models, internet hosts grouping method and the evaluation model were focused on. A new internet hosts grouping algorithm called Clustyes was presented. Also, based on the concept of group coupling degree, an evaluation model of hosts grouping performance was constructed. RTT, hops, and packet loss rate in network communication were considered in this evaluation model. The experiment was simulated on the NS-3, and the results show that the new grouping algorithm Clustyes can generate groups, in which the hosts can communicate quickly, stably, and closely in physical distance.

关 键 词：分组评价模型 组耦合度 聚类分析 贝叶斯分类器 

分 类 号：TN918[电子电信—通信与信息系统]