This paper presents a new airport clustering algorithm that facilitates the design of an air traffic flow network in support of a Flow Contingency Management decision support system concept. As Flow Contingency Management is concerned with decision making in the strategic timeframe, aggregating airports into clusters provides a useful mechanism to capture the traffic while limiting network size; however the specific needs of the Flow Contingency Management Framework require an alternate approach to airport clustering. To meet these needs, the algorithm proposed, termed Split by City-Pair, first assigns all eligible airports to one cluster and then, based on origin-destination city pair information, uses a hierarchical top-down method to split clusters until no cluster has a top city-pair contained within. Four metrics were defined to evaluate the quality of the clusters generated using the proposed method as well as two well-known clustering algorithms, the K-means method and the Weighted Proximity Classifier method. The results show that the Split by City-pair method performs as well or better than the other two methods while also satisfying the specific requirements of a Flow Contingency Management network.
展开▼
