Prophet混合模型应用于基站网络流量长期预测

传统网络流量预测方法大多数关注短期预测，而长期预测能够更好地指导基站小区无线设备扩缩容。集合经验模态分解（ensemble empirical mode decomposition，EEMD）能够使非平稳时间序列转化成平稳时间序列，Prophet模型能够准确地对流量序列进行较准确的长期预测，基于以上模型方法的优点和基站小区网络流量的非线性和非平稳性特点，提出一种Prophet混合EEMD的基站小区网络流量预测方法（E-Prophet）。采用EEMD将网络流量序列分解成若干固有模态函数（intrinsic mode functions，IMF）分量和一个残差分量；利用Prophet模型对各分量建模，并将各分量预测结果进行线性组合，得到最终的预测结果。利用实际基站小区网络流量数据对方法进行验证，结果表明：E-Prophet对于网络流量长期预测比Prophet、SARIMA、LSTM以及结合EMD和Prophet的模型具有更高的准确度和鲁棒性。

Application of Prophet Mixture Model on Long-Term Prediction of Base Station Cell Network Traffic

Most of the traditional network traffic prediction methods focus on short-term, while long-term prediction can better guide the base station cell wireless equipment expansion and contraction. Ensemble empirical mode decomposition（EEMD） can convert the non-stationary time series into a stationary time series, and Prophet model can accurately make a more accurate long-term prediction of traffic series. Based on the advantages of the above model methods and the non-linear and non-stationary characteristics of the base station cell network traffic, Prophet hybrid EEMD method（E-Prophet） is proposed for base station cell network traffic prediction. First, the EEMD is adopted to decompose the network traffic series into several intrinsic mode functions（IMF） components and a residual component; then, the Prophet model is used to model each component, and the predictions of each component are linearly combined to obtain the final prediction result. The method is verified using real-world base station cell network traffic data, and the results show that E-Prophet model has higher accuracy and robustness in long-term prediction of network traffic compared with Prophet, SARIMA, LSTM, and the model combining EMD and Prophet.