Fig 1.
The architecture of the agent-based flight delay prediction model.
Table 1.
Notations and descriptions.
Fig 2.
The busiest 34 airports in the United States of America.
The map was created by Python 3.6 with the help of free vector and raster map data from Natural Earth (http://www.naturalearthdata.com/) and the location data of airports from BTS (https://www.bts.gov/).
Fig 3.
Histogram of the frequency distribution of the departure delay and the probability density function of the departure delays.
Fig 4.
The Q-Q plot.
Fig 5.
Random Forest regression (RFr) method performance with the estimated elapsed time (EET) and minimum turnaround time (MTT) models for a different number of trees.
Table 2.
Flight information and delay levels on test days.
Fig 6.
(a) indicates MAE of the predicted delays of thirty test days and (b) indicates RMSE of the predicted delays of thirty test days.
Table 3.
The performance for three types of test days and all test days with different measurements.
Fig 7.
The empirical cumulative distribution function (ecdf) of the absolute errors of flight delays.
Fig 8.
The contribution of the parameter methods.
Fig 9.
Forecast horizon vs. MAE.
Table 4.
The performance of existing literature.