Air Traffic Flow Forecasting Using Multi-Feature Elman Neural Network

In air traffic flow management, air traffic flow forecasting is the most important part; it not only can improve security and effective utilization of airspace and airport resources but also can greatly improve airlines' economy benefits and operational efficiency. In this paper, the Elman neural network prediction method, which achieved relatively good results in prediction of ground traffic, is applied in air traffic flow prediction, and a new air traffic flow prediction method based on multi-feature Elman neural network is proposed. First, the velocity characterized by the first derivative and the acceleration characterized by the second derivative are introduced as two new features into the structure of the single-feature Elman neural network, and a multi-feature Elman network is built. Further, the parameters of the network structure are studied by using the steepest descent method with the driving quantity items. The air traffic flow of 36 cities in East China are tested as the experimental data of the proposed method. Experimental results show that multi-feature Elman method compared with single-feature Elman method can obtain better prediction results.