论文部分内容阅读
由于网络环境的动态性及网络应用分布变化等原因,导致在P2P流量识别过程中面临概念漂移问题.现有的静态流量识别模型无法处理此类问题,因此文中提出一种新的P2P流量识别模型,解决了P2P流量识别过程中存在噪声和概念漂移问题.借助聚类思想利用K近邻算法,实现噪声过滤.根据评估假设原理和中心极限定理实现了概念漂移检测,在此基础上利用基分类器的不确定性输出结果实现多分类器集成方案,同时基于时间策略淘汰概念过时的基分类器.理论分析和仿真实验结果证明了本文所提的算法模型是可行的.
Because of the dynamic network environment and the change of distribution of network applications, it causes the concept of drift in the P2P traffic identification process.The existing static traffic identification model can not handle such problems, so a new P2P traffic identification model , Which solves the problem of noise and concept drift in P2P traffic identification process.K-nearest neighbor algorithm is used to realize noise filtering based on clustering idea.According to the evaluation hypothesis principle and the central limit theorem, the concept drift detection is implemented, and based on this, The uncertainty of the output results to achieve multi-classifier integration scheme, and outdated base classifier obsolete based on time strategy.Theoretical analysis and simulation results show that the proposed algorithm model is feasible.