基于小波分析和神经网络的网络流量预测

Prediction of Network Flow Based on Wavelet Analysis and NN

采用小波分析和神经网络工具对分时段网络流量进行预测,比基于顺序流量序列的预测方法具有更高的预测精度。首先将分时段网络流量序列进行小波分解后得到的各子序列分别用神经网络进行训练,然后将各子序列预测结果进行重构作为最终的预测结果。文章最后将不同的小波分解和分解水平的预测结果误差作了比较,指出应根据实际的网络流量序列的变化规律选择合适的小波;小波分解水平不宜过高,以避免重构误差的累加。

The model, which adopts wavelet analysis and neural networks to predict network flow based on period of time, has higher prediction precision than the prediction way based on sequence flow series, Each sub-sequence from network flow based on period of time analyzed by wavelet is first trained by neural networks separately. Then prediction result of each sub-sequence is rebuilt and taken as the final prediction result. Finally this paper points out that the proper wavelet should be chosen according to real network flow through the comparison of prediction results be too high, thus avo by idi different wavelet analysis and analysis level, and that the analysis level should not ng error accumulation of rebuilding