Research On Departure Flight Delay Prediction Model And Algorithm Considering Air Route Network Structure

With the rapid growth of air traffic year by year,the limited airspace resources and ground multiparty security resources are difficult to meet the increasing air traffic flow,which aggravates the occurrence of flight delays.Frequent flight delays not only seriously affect passengers’ travel plans and reduce their satisfaction with civil aviation services,but also increase the workload of airlines,airports,air traffic control and other units,and cause economic losses,which hinder the safe and efficient development of civil aviation industry.Through effective technical methods,it is of great practical significance to make accurate prediction of flight delays in advance,so as to provide decision-making reference for the relevant departments of civil aviation to make countermeasures in advance and reduce the negative impact of flight delays.Flight delays are caused by many complex factors,and the current flight delay prediction lacks the consideration of air route network structure factors,and the accuracy of traditional multi-category prediction is not high.Therefore,in this paper,a departure flight delay prediction model considering the air route network structure is constructed,and the deep learning optimization algorithms are designed for the departure flight delay prediction model,aiming to effectively improve the prediction accuracy of the model and provide a reference for the decision making of civil aviation units.The main research contents are as follows:(1)The air route network structure of the terminal area where the departure flight is located is quantitatively analyzed.Firstly,the air route network structure corresponding to the terminal area of the airport where the departure flight is located is selected as the research object,and the analysis is carried out in two dimensions: air route network and network structure.Secondly,three types of air route congestion indexes are proposed,namely air route flow,air route congestion degree,and air route network congestion degree.Finally,based on these congestion indexes,an example is quantified to analyze the flight delay characteristics,air route and air route network congestion characteristics in the terminal area.(2)A departure flight delay prediction model considering the air route network structure is constructed by using deep neural network.Firstly,the collected flight information data,meteorological information data and quantified air route congestion data are pre-processed.Secondly,the departure flight delay prediction model considering the air route network structure is constructed by using deep neural network.Finally,by adjusting the Focal Loss function as well as the hyperparameters,numerical experiments with different loss functions and different data sets are carried out.The experimental results show that the model prediction accuracy reaches 93.47%,which verifies the effectiveness of the air route network structure factor.(3)The optimization algorithms of the departure flight delay prediction model are designed.Firstly,several current typical deep learning optimization algorithms are sorted and analyzed to compare the iterative characteristics and performance effects of different algorithms.Secondly,two optimization algorithms are designed on this basis: the Accelerating Nadam algorithm and the Hybrid Nadam algorithm.Finally,based on the constructed departure flight delay prediction model,numerical experiments of different algorithms are carried out,and the experimental results show that the model prediction accuracy of both algorithms is improved compared with the traditional Nadam algorithm,reaching up to93.69%,which verifies the effectiveness and usability of the algorithms to improve the model prediction accuracy.