|
Aiming at the problems of low-efficiency issues and flight delays caused by the unanticipated taxi-out time of departure aircraft at mega-airports, a departure aircraft taxi-out time prediction model was constructed based on deep feedforward networks. Firstly, the relevant factors affecting the departure taxiing of flights were analyzed, then the entity embedding method was used to encode categorical variables, which transformed categorical variables into a set of vectors by training neural networks, and numerical variables were processed by standardizing, then all these variables were concatenated and used as input to the prediction model; subsequently, a deep feedforward neural network model was developed for predicting the taxi-out time of departure flights; finally, the historical flight operation data of Kunming Changshui International Airport was used as an example for validation. The prediction results show that compared with traditional machine learning algorithms, the average prediction accuracy of the proposed model is up to 95.8% within the error range of ±5minutes, meanwhile, it is shown that the adopted entity embedding method encoding categorical variables is superior to one-hot encoding.
|
