Deep learning-based prediction of base station traffic

Nowadays, the development of 5G, edge computing, NFV and other technologies brought by surge network traffic will become a new challenge to refinement, automation, intelligent operation maintenance management network. In order meet this challenge, it is necessary accurately perceive application-level at multiple levels, such as network, MAN backbone reduce error predicting flow data, neural algorithm prediction model based on machine deep learning, long short memory model, which can predict base station data according periodicity volatility characteristics data. After experimental verification, shows that compared with traditional time series AR ARIMA also has basic is, fully connected model. This method higher accuracy smaller in mobile communication prediction. The MAE value optimized 21.6%, 33.4% 12.5%.