海量非线性网络流量数据分类模型的仿真分析

Simulation Analysis of Nonlinear Classification Model for Huge Amounts of Network Traffic Data

当海量非线性网络流量数据分类过程中的数据支持量增多时,采用传统的方法组建的流量分类模型,对其具有的二重性和混沌特性考虑不充分,建模时需要大量的迭代计算,分类器训练过于缓慢,导致分类精确度低,抗噪性差。提出基于改进SVM算法的海量非线性网络流量数据分类方法。针对海量非线性网络流量数据分类的初始训练样本集进行聚集和压缩,重新建立新非线性网络流量数据分类样本集,利用SVM算法训练流量数据分类器,在依据不同的流量数据分类样本对每个流量类型特征贡献的大小给定相应的模糊因子,建立流量数据分类模型,引入网络流量分类特征有效度,构造其特征有效度核函数,对上述分类模型的分类精度进行优化。仿真结果证明,改进SVM算法的海量非线性网络流量数据分类方法分类精确度高,抗噪性能良好。

A classification method of massive nonlinear network flow data is presented based on improved SVM al- gorithm. The initial training sample set of nonlinear network flow data classification is aggregated and compressed, and a new nonlinear network flow data classification sample set is established. The traffic data classifier is trained by using the SVM algorithm. The corresponding fuzzy factor is obtained according to the contribution of different traffic data classification samples to each flow type trait, and a traffic data classification model is established, the characteristic effective degree of the network traffic classification is introduced, the kernel function of the characteristic effective degree is constructed, and the classification accuracy of the classification model is optimized. The simulation shows that the classification method based on the improved SVM algorithm has high classification precision and good antinoise performance.