一种主动学习式P2P流识别方法

P2P流的识别对于网络的维护与运营都具有重要意义,基于机器学习的流识别技术是目前研究的热点和难点内容,但目前仍然存在着建立分类模型需要大量适用的训练数据、训练数据的标记需要依赖领域专家以及因此而导致的工作量及难度过大和实用性不强等问题,而当前的研究工作很少涉及到这些问题的解决办法。针对这一问题,采用主动学习技术提取少量高质量的训练样本进行建模,并结合SVM分类算法提出了一种基于锦标赛选择的样本筛选方法。实验结果表明,其相对于已有的流识别方法,能够在仅依赖少量高质量训练样本的前提下,保证较高召回率及较低误报率,更适用于现实网络环境。

Peer-to-peer traffic identification using active learning

The identification of peer-to-peer(P2P) traffic is essential to network maintenance and management.Traffic identification using machine learning(ML) techniques has been an active and difficult research topic for recent years.However,there are still some unresolved and scarcely addressed problems such as the difficulties in obtaining adequate qualified training data for the supervised classifiers to model traffic patterns,the data acquisition task is always time-consuming and greatly relies on the domain experts.To solve the problem,this paper introduced active learning method to select the most qualified data for training and proposed a tournament selection based instance selection method for SVM.The experimental results demonstrate that the proposed method is able to guarantee high recall rate and low false positives by using a small quantity of high qualified data.Therefore,it is more suitable for the real network applications than the traditional ones.