Research On The Aircraft Pushback Time Decision Based On Machine Learning

The current release mode of determining pushback time by extensively estimating aircraft taxi-out time causes the low operation efficiency and the low runway utilization efficiency.Thus it is urgent to find a precise method to predict taxi-out time.The taxi-out time of departures is influenced by many external factors,and with the advent of Big Data Era,the traditional prediction methods are no longer applicable.Forfunately Machine Learning come into being.It can mine rules from historical data and predict the operation on the surface in future period.Based on this method,taxi-out time can be predicted precisely.Finally,this paper can provide technical support for aircraft pushback time decision.The main contents and results are as follows:(1)Based on K-Nearest Neighbor(KNN),a classification model is estabilished to predict the taxiing route.In order to improve KNN search efficiency,an improved search algorithm based on KD Tree is proposed.Then this paper designs an aircraft taxiing route prediction algorithm based on searchimproved KNN.The experimental results show that the prediction accuracy is up to 97.87%.Based on the predicted taxiing route,the taxiing distance of each flight is calculated.(2)Considering the prediction inaccuracy caused by traditional Naive Bayesian Classifier due to independence assumpations between attributes,KDE method is introduced and Non-Naive Bayesian Classifier based on joint probability density function estimation is proposed.After building a feature variable set,a prediction model for the number of departures is established based on Non-Naive Bayesian Classifier.The experimental results based on the actual operation data of Shanghai Pudong Airport indicate that the prediction accuracy of the proposed model is up to 94.21% within ?4.(3)According to all the influencing factors such as predicted taxiing routes and the number of departures,this paper builds the optimal feature variable set by feature selection based on Gradient Boosting Tree(GBT).Then,a taxi-out time regression prediction model is constructed based on Gradient Boosting Regression Tree(GBRT).The simulation which takes Shanghai Pudong Airport as an example shows that the prediction accuracy is up to 94.31% within ?5min,and verifies that the prediction effect is better than other regression methods.Combined with the predicted taxi-out time and planned take-off time,the optimal aircraft pushback time is deprived.The research results in this paper can provide precise taxi-out time and optimal pushback time,which use scientific method to assist air traffic control,improve flight release rate and runway utilization efficiency,provide technical support for precise control decision and optimizing control strategy.