Mining aviation data to understand impacts of severe weather on airspace system performance

This paper describes our latest experiment with application of data mining to analyzing severe weather impacts on National Airspace System (NAS) performance. We show the importance of data preparation and feature extraction in our work. Two types of data - weather and air traffic data - were used in this experiment. Weather data are represented as binary images. A severe-weather day for air traffic is represented as a set of severe-weather regions, each with a set of weather- and traffic-related features. The set of severe-weather regions for each day was first converted into a vector of attribute values, and then classification, regression and clustering were applied to the data. Initial results were encouraging, while later results were improved and impressive. Meaningful classification rules were generated and the clusters generated for weather-traffic days were clearly correlated with NAS performance.