融合模拟退火算法的神经网络流量预测研究

Traffic prediction research of neural network combined simulated annealing algorithm

为了提高网络流量预测的精度,针对BP网收敛极易陷入局部极小点的缺陷,引入模拟退火算法思想优化小波包神经网络,对网络流量数据的时间序列进行建模预测。先将原始网络流量序列进行小波包消噪,将消噪后的序列作为融合模拟退火思想的小波包神经网络的输入,待预测序列作为输出。通过消噪后的前N天的流量序列,预测出后M天流量序列。仿真实验结果表明,与直接利用小波神经网络预测的模型比较,融合了模拟退火算法思想的小波包神经网络具有更好的预测能力。

In order to improve the precision of the network traffic prediction, measures should be taken to solve the problem that the BP network prone to the local minimum, annealing algorithm is used to optimize the BP network to model and predict the time series of network traffic data. The processing procedure is, firstly, denoise the traffic time series with wavelet packet transform, then taking the denoised series as the input of the BP wavelet packet neural network combined with simulated annealing, while the predictive series as the output of neural network. Using the earlier N day＇ denoised traffic time series to forecast the later M days＇ predictive series. The experimental results prove that, compared the method only using wavelet neural network with the model based on idea of simulated annealing show that the model have better predictive ability.