一种基于机器学习的P2P网络流量识别算法研究  

Study on P2P Network Flow Recognition Algorithm Based on Machine Learning

作　　者：袁华兵 YUAN Huabing(Information Technology Department,Xi'an Medical College,Xi'an 710021)

机构地区：[1]西安医学院信息技术处

出　　处：《计算机与数字工程》2019年第10期2387-2391,共5页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号：61634004)资助

摘　　要：为提高P2P网络流量识别的准确率,针对Elman神经网络预测精度受其权值和阈值选择的影响,论文运用蜻蜓算法对Elman神经网络的权值和阈值进行优化选择,提出一种基于DA-Elman的机器学习的P2P网络流量识别模型。将TCP流量比例、连接数与不同IP数目的比值、平均数据包长度、上行流量比例、数据包总数5个特征属性作为DA-Elman模型的输入,网络流量类型作为DA-Elman的输出。与PSO-Elman、GA-Elman和Elman相比,研究结果表明,DA-Elman可以有效提高P2P网络流量识别的准确率,其准确率高达98.4252%,为P2P网络流量的识别提供新的方法和途径。In order to improve the accuracy of P2P network traffic recognition,this paper uses the dragonfly algorithm to optimize the weight and threshold of Elman neural network.A P2P network traffic recognition model based on DA-Elman machine learning is proposed.The five characteristic attributes of the TCP traffic ratio,the number of connections to different IP numbers,the average packet length,the uplink traffic ratio,and the total number of packets are used as inputs to the DA-Elman model,and the network traffic type is used as the DA-Elman output.Compared with PSO-Elman,GA-Elman,and Elman,the research results show that DA-Elman can effectively improve the accuracy of P2P network traffic recognition,with an accuracy rate of 98.4252%,providing new methods and approaches for P2P network traffic identification.

关 键 词：机器学习 ELMAN神经网络 粒子群算法 遗传算法 网络流量识别 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]