大规模混合网络中基于朴素贝叶斯分类的TCP自适应鉴别器  

TCP adaptive discriminator based on naive Bayes classifier in large-scale hybrid networks

作　　者：陈娅婷 鲁凌云[1] CHNE Yating;LU Lingyun(School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China)

机构地区：[1]北京交通大学计算机与信息技术学院,北京100044

出　　处：《通信学报》2018年第A01期189-194,共6页Journal on Communications

基　　金：国家自然科学基金资助项目(No.61771002);赛尔网络下一代互联网技术创新项目(No.NGII20170636)~~

摘　　要：针对有效区分分组丢失类别问题,提出了一种端到端多参数度量自适应拥塞控制算法——TCP-NBayes算法。首先根据混合网络下有线链路和无线链路分组丢失特性的不同选取特征参数。然后基于朴素贝叶斯理论建立区分无线随机丢失和拥塞丢失的鉴别模型。仿真实验表明该算法分类准确度最高可达95%,与其他算法相比,其友好性与公平性表现良好,对网络性能有很大的改善。With the rapid integration of wireless technology and wired networks,the increasing user service requirements for network performance are also increasing.The traditional TCP protocol attributes frequent random packet loss in the network to congestion,which greatly reduces network performance.Aiming at effectively distinguishing the packet loss category and improving the network transmission performance problem,an end-to-end multi-parameter metric adaptive congestion control strategy is proposed.TCP-NBayes,based on the naive Bayesian theory in machine learning,establishes the distinction between wireless random loss and Identification model for congestion loss.Simulation experiments show that the classification accuracy of this method is up to 95%.Compared with other algorithms,the friendliness and fairness are good,and the network performance is greatly improved.

关 键 词：无线/有线混合网络 TCP 朴素贝叶斯分类 机器学习 

分 类 号：TN913.2[电子电信—通信与信息系统]