Time Series Analysis And Prediction Of Airport Flight Delay

With the increasing volume of civil aviation flights,flight delay is a difficult problem and challenge that civil aviation transportation must face.It is easy to conduct post analysis of factors affecting flight delays.But under the premise that the factors affecting flight delays are difficult to predict,how to realize accurate and effective prediction of flight delay conditions in the future by learning the change trends or behavior patterns of the time series of delay conditions is an urgent need to overcome in the process of development of civil aviation.In this regard,this article combines data from an airport for five years from 2014.1.1 to 2018.12.31 to carry out research.First of all,the flight delay condition is calculated from three aspects: delayed flights,rate of delayed flights and average delay time,and time series of flight delay condition is constructed.Secondly,the nonlinear characteristics of the time series of delay conditions are studied,and the chaos and fractal characteristics of three types of time series of delay conditions are discussed.After that,the predictability of different time series is qualitatively analyzed using Recursive Graphs.The results show that the time series of arrivals,departures,delays,delays,and average delays have chaos characteristics.R/S analysis shows that the sequences have long memory,future trends are positively correlated with the past,and the delay of departure flights at any time can affect the delay condition in the next 15 hours.Thirdly,a chaos prediction model of the time series of delay conditions is constructed based on the Extreme Learning Machine(ELM).Analysis of data examples shows that the model has a good prediction effect on flight delays,with a Mean Absolute Error of approximately 3.5 flights.Finally,the Ensemble Learning model based on Stacking is used to build a flight delay level early-warning model.The model integrates four machine learning algorithms: Random Forest,Support Vector Machine,Neural Network and XGBoost.The analysis of the prediction accuracy shows that the prediction results of the Ensemble Learning model are better than that of each single model,and the prediction accuracy is 81.82%.The prediction results have substantial value for the flight delay warning and response.Based on the time series of flight delays,this paper proposes a chaos prediction model of flight time delays based on ELM on the basis of identifying the chaos characteristics of the time series,and proposes a stacking-based Ensemble Learning model about flight delay level early warning based on four machine learning algorithms.These models are of great significance to the prediction and early warning of flight delays in an airport.