基于Kohonen神经网络的组合式流量预测模型

Combined Prediction Model of Network Traffic Based on Kohonen Neural Network

摘要: 该文针对传统预测模型预测精度低、对训练数据依赖程度高以及不能很好的刻画网络流量特征等不足，提出了一个混合的流量预测模型。该模型根据Kohonen神经网络的学习速率快、分类精度高、抗噪声能力强等特性，通过小波变换将网络流量分解为高频部分和低频部分，高频部分采用Kohonen神经网络进行预测，低频部分采用自回归AR模型进行预测，并采用Matlab进行仿真，通过实验得，这种组合预测模型可以提高对非线性、多时间尺度变化的网络流量的预测精度。

Abstract: Considering that the original prediction model whose accuracy is low, and which highly depends on the training data and can’t well described the characteristics of network traffic, we proposed a mixed traffic prediction model. The model is based on the Kohonen neural network feartures, that is, quickly learning rate, highly classification accuracy and strongly anti-noise. By wavelet trans-forming, we decompose the network traffic into high frequency part and the low frequency part, and the high frequency part is dealt by using Kohonen neural network prediction model, the low frequency part by using autoregressive AR model to predict by using Matlab to simulat. Through the experiment we conclude this combination prediction model can improve the prediction accu-racy on multiple time scales and the nonlinear changing network traffic.