Short-term Flight Trajectory Prediction Based on LSTM-ARIMA Model

The efficient and accurate trajectory prediction is the key technologies of future air traffic management system, which aims to improve the operational capability and the predictability of air traffic. Aiming at the problems of insufficient accuracy and instability of existing prediction methods, some new dimension features are constructed based on the existing historical trajectory data. By analyzing the statistical characteristics of three-dimensional data of longitude, latitude and height, a combined prediction model based on LSTM and ARIMA is proposed, which combines LSTM network's strong approximation ability to non-linear and non-stationary time series and ARIMA's better prediction performance in linear time series. Firstly, LSTM is used as the main prediction model to predict longitude, latitude and height, and then ARIMA is used to model linear relationship of height. Finally, CRITIC method is used to fuse the height values of LSTM and ARIMA prediction. The experimental results show that the combined model takes advantages of two models and improves the accuracy of trajectory prediction.