Classification of flight takeoff risk level based on selective ensemble clustering

Abstract

Takeoff is the most accident-prone phase of flight operation. It is necessary to assess the takeoff risk levels for every flight according to the operation conditions to avoid accidents. However, there is no recognized assessment standard for takeoff risk levels in the field. In this paper, we propose a takeoff risk level classification method based on selective ensemble clustering. First, we use multiple heterogeneous base clusterers to cluster the flight takeoff data respectively and obtain different cluster division results. Then, the selective cluster ensemble method is used to relabel the cluster labels generated by all base clusterers, so that similar clusters have the same label. Finally, these calibrated cluster markers are used to generate the final ensemble clustering result, i.e., the classification of risk levels. Experiments on actual operating data sets show that the proposed flight takeoff risk classification result based on selective ensemble clustering is more general. The assessment performance is better than that of a single clusterer.