Go-Around Prediction Modelling: A Comparison of Two Airports

Abstract

In this paper, we employ the Long Short-Term Memory model (LSTM) to predict the real-time probability of a go-around as an arrival flight approaches an airport within 10 nautical miles of the landing runway threshold. We train the model on two airports–New York JFK and San Francisco SFO. We further develop methods to analyze the causes of go-arounds. Our results indicate that the models for both airports perform quite well and that risk of simultaneous runway occupancy is the primary factor contributing to overall go-around occurrences. The models will be integrated with real-time data streaming and a web-based user interface to provide a real-time alerting capability that can be used by flight crews and other line personnel to identify and manage situations with a high risk of a go-around.