基于小波分解的网络流量模型

论文充分利用小波变换具有多分辨率的特点,将时域里的网络流量通过小波分解,分解到不同的频带上。再对各子频带上的细节分量使用不同阈值进行消噪处理,使分解后的流量在频率成分上较单一,且平稳性较好。然后采用自回归滑动平均混合模型对小波分解去噪后的不同分量分别进行预测再合成预测流量。对实际流量进行模拟预测,结果表明该模型有效地提高了预测精度,能对网络流量特别是短期流量作出较为准确的预测。

Network Traffic Model Based on Wavelet Decomposition

Through making full use of the multiresolution analysis in the wavelet transform,the network traffic,which is difficult to analyze and model in the time domain,is divided into different bands of frequency by using wavelet decomposition.Then according to different sub-bands of frequency,different thresholds are adopted to carry on the denoising disposition,which can simplify the frequencies of the decomposed traffic and smooth the traffic.Auto Regressive Moving Average Model is employed to predict the denoised sub-traffics with different bands of frequency.Then the expected traffic to predict in time domain is obtained by synthesizing these sub-traffics.Experimental prediction result with the real network traffic indicates that the model improves the prediction accuracy and provides a relatively accurate forecast of the network traffic,especially the short-range traffic.