Communication Base Station Traffic Prediction Based on Nonlinear Regression-LSTM Decomposition Algorithm

Communication base station traffic prediction can help network service providers predict the peak hours of network traffic, so as to carry out network expansion and load balancing in advance, and keep the users getting a smooth network experience even during the peak hours of the network. In order to accurately predict traffic usage in different time periods, we first constructed a nonlinear regression model, then solved the parameters in the algorithm, and finally got the image of our predicted nonlinear function and compared it with the real value, and found that there are some small swings. Finally, we use the LSTM recurrent neural network to predict the swing term. The final flow prediction is obtained which is very close to the real situation. The model performance was verified with MAE, MAPE and RMSE metrics respectively, and the results showed that the values of the three metrics were 0.1873, 0.0978 and 0.2260 respectively.