Ground stop analysis & prediction

• The EWR GSs were enacted reactively to a sudden/unexpected imbalance of airport demand and capacity and used to preclude extended airborne holdings. Sometimes, the actual GS durations were extended from the planned ones up to 3 hours or even longer (4%). Multiple GSs were enacted in 25% of days investigated. TFM made a TMI transition from GS into a GDP in 13% of days at the EWR airport. • Over the years 2007–2009, 54% of the days had a GDP implemented. GSs were enacted during 65% of the GDP days, (for 40% of the GDP days, GS enacted during the GDP; for 25%, TFM made a TMI transition from a GS into a GDP event). • The GS predictions are accomplished by using BDT. The supervised machine learning is employed to train the models. The models are validated using data cross validation methods. • When predicting the occurrence of GS, GDP, and GS/GDP from the normal days, the model was able to achieve an overall accuracy rate about 85%. In the study to distinguish the GS/GDP days from GDP/Non-GS days an overall accuracy rate of 71% was achieved.