Research on Air Traffic Flow Management Delay Distribution Prediction Based on IV Value and PSO-SVM

Abstract

Air Traffic Flow Management delay (ATFM delay) can quantitatively reflect the spatial and temporal distribution of congestion caused by the imbalance of airspace network capacity and demand, and is an important parameter to evaluate the effectiveness of air traffic flow management (ATFM) strategy. If the distribution of ATFM delay can be predicted in advance, the predictability and effectiveness of ATFM strategy can be improved. In this paper, we propose a short-time ATFM delay spatial and temporal distribution prediction method in the absence of historical records of specific ATFM delay causes. The method first clarifies the prediction object by constructing ATFM delay prediction network model, and proposes the ATFM delay prediction index system; Secondly, we use Information Value (IV) to filter and weight the features in order to construct a weighted ATFM delay prediction index system; Then, a support vector machine model based on particle swarm optimization(PSO-SVM) is implemented to predict ATFM delay distribution. Finally, some congested airports and key waypoints in China are selected as ATFM delay prediction network nodes for empirical analysis. The experimental results show that the prediction accuracy of the IV-PSO-SVM model reaches 96.4%, which is 13.5% and 9.1% higher than that of SVM model and PSO-SVM model, respectively.

Keywords

• ATFM delay
• delay prediction
• airspace network
• IV value
• PSO
• SVM