Data mining for robust flight scheduling

Abstract

In scheduling of airport operations the unreliability of flight arrivals is a serious challenge. Robustness with respect to flight delay is incorporated into recent scheduling techniques. To refine proactive scheduling, we propose classification of flights into delay categories. Our method is based on archived data at major airports in current flight information systems. Classification in this scenario is hindered by the large number of attributes, that might occlude the dominant patterns of flight delays. As not all of these attributes are equally relevant for different patterns, global dimensionality reduction methods are not appropriate.We therefore present a technique which identifies locally relevant attributes for the classification into flight delay categories. We give an algorithm that efficiently identifies relevant attributes. Our experimental evaluation demonstrates that our technique is capable of detection relevant patterns useful for flight delay classification. © 2009 Springer US.