一种基于BP网络的流量分类方法

Traffic Classification Method Based on BP Network

针对传统的网络流量分类方法准确率低、开销大、应用范围受限等问题,提出了一种基于BP网络的流量分类方法。该方法改进了标准的BP网络算法,采用基于Lyapunov函数得到的自适应学习率,并引入遗传算法优化网络的初始连接权值和阈值,使网络避免陷入局部最小,加速了网络收敛过程。实验结果表明,采用改进的BP网络算法来处理网络流量分类问题具有明显的优势:该方法的收敛速度和拟合精度均优于标准BP算法,而且流量分类准确率高于NB算法。

In order to solve low accuracy and limited application region in traditional traffic classification,a method of network traffic classification based on BP network is proposed.This method improves traditional BP algorithm,and applies the adaptive learning rate produced by Lyapunov function.Genetic algorithm is adopted in the connection power and threshold of network optimization to avoid local minima of network,but accelerate network convergence speed.The experimental results show that the convergence speed is faster with a high fitting precision than traditional BP algorithm,and a higher network traffic classification precision than NB algorithm.