论文部分内容阅读
为了更好地对网络流量进行分析和管理,提出一种基于小波变换、自回归滑动平均模型(ARMA)和极限学习机(ELM)的组合预测模型W-ARMA-ELM.原始数据通过小波分解产生近似序列和细节序列,通过对分解序列的自相关性和偏自相关分析,平稳序列使用ARMA预测,而非平稳序列使用ELM预测.使用兰州大学教育网、网通流量数据和英国学术主干网流量数据三组不同的网络流量数据来检验组合模型W-ARMAELM的预测性能.实验结果表明提出的组合方法要比单一的ARMA和ELM预测效果要好.同时指出使用自相关和偏自相关分析相结合的方法对分解后的子序列进行平稳性判定有助于选择合适的组合模型从而提高预测精度.
In order to analyze and manage the network traffic better, a combined forecasting model W-ARMA-ELM based on wavelet transform, autoregressive moving average model (ARMA) and extreme learning machine (ELM) is proposed. The original data is generated by wavelet decomposition Approximate sequence and detail sequence.According to the autocorrelation and partial autocorrelation analysis of the decomposed sequences, the stationary sequence is predicted by ARMA, while the non-stationary sequence is predicted by ELM.Using Lanzhou University education net, Netcom flow data and UK academic backbone flow data Three different sets of network traffic data to test the performance of the combined model W-ARMAELM.Experimental results show that the proposed combination method is better than the single ARMA and ELM prediction results.At the same time, the combination of autocorrelation and partial autocorrelation analysis Determining the stability of the decomposed subsequences helps to select the appropriate combination model to improve the prediction accuracy.