Experience with a System for Manual Clustering of Air Surveillance Track Data
Matthew C Lowry · 2013
Abstract : Using clustering algorithms to mine air surveillance track data for groups of similar flights has the potential to facilitate a variety of capability enhancements. Since there are many algorithms that could be used, a method for assessing the quality of algorithm output is required. One potential method is to have a human expert hand-craft a clustering for a test data set, and use this manual clustering as the gold-standard against which the output of a clustering algorithm is assessed. For complex spatio-temporal data such as air surveillance track data, the manual construction of clusterings for a robust test data suite will be labour-intensive and reliant on good information technology support. This report describes an experimental system providing a user interface and workflow for performing manual clustering of air surveillance track data, and experience with a trial of the system.