Traffic Prediction Method for Time Series Networks Based on ARIMA-LSTM Model

The increasing scale of network access in recent years requires an advanced analysis of future network data traffic(NDT) trends to anticipate various network scenarios. However, current prediction methods are limited in accuracy due to the overly complex components of network traffic on the time scale. This paper proposes a time series network traffic prediction(NTP) method based on the ARIMA-LSTM model. The method uses the ARIMA to extract the linear components of the web traffic data and then uses the LSTM to extract the nonlinear components further. It will achieve effective prediction of complex time series. The article validates the effectiveness of the method experimentally on two publicly available datasets. All experimental results show that the ARIMA-LSTM method has satisfactory prediction performance and outperforms several other state-of-the-art works. In sum, this NTP method points to a new orientation for NDT monitoring and analysis.