基于Spark的网络流量分类方法研究

Network traffic classification based on Spark frame

网络流量分类是网络监控、服务质量和网络安全的前提和基础。为了对海量网络流量进行迅速、准确地分类,利用相似性和权重对随机森林算法进行改进,通过计算决策树相似度,消除决策冗余以加强分类性能;再用分类性能为指标设定权重构建随机森林,并结合Spark平台设计实现并行算法提高分类效率。实验结果表明,该方法提高了网络流量分类性能,同时具有可扩展性和顽健性,能够响应海量流量分类任务。

Network traffic classification(NTC)is the premise and foundation of detecting network monitoring,Quality-of-Service(QoS)management and network security.In order to classify the vast network traffic quickly and accurately,the random forest classification algorithm based on similarity and weights was improved,which eliminate decision-making redundancy through the similarity.Then classification performance index enactment right heavy was used to build a random forest.Combined with Spark platform design and implement parallel algorithm.The experiments show that our method improves the classification accuracy,has scalability and robustness,and can respond to the massive traffic classification task.