Design And Implementation Of Flight Delay Prediction System Based On Flight Chain

In recent years,with the rapid growth of the national economy,more and more people choose more convenient ways to travel,especially by air,and with this comes a huge increase in the number of flights.The problems caused by flight delays have greatly affected the travel of passengers,who are eager to be informed of the exact time of flight departure or arrival so that they can make reasonable arrangements for their next travel plans.At present,most of the research has been conducted on the classification problem of flight delays and give the prediction result of whether a flight is delayed or not;for the regression problem of flight delays,the research has only been conducted on one aspect of flight departure or arrival delays,and in the process of research,the prediction results are not satisfactory due to the many and complex factors affecting flight delays,the interaction between flights and flights,and most of them are single models.To address the above issues,this paper investigates the flight delay prediction problem as follows:(1)Considering the many factors affecting flight delays and the highly nonlinear relationships among them,we propose a Support Vector Machine(SVM)flight departure delay prediction model based on the Improved FA Algorithm(IFA).The SVM is used for flight departure delay prediction,and then the proposed IFA algorithm is used for support vector parameter finding to construct the flight delay prediction model.By testing the model against the U.S.public flight dataset,the results show that the accuracy predicted by this model’s has been improved.(2)Considering the problem that a single model cannot predict accurately,a flight arrival delay prediction model based on a weighted mixture of Long Short-Term Memory(LSTM)and eXtreme Gradient Boosting(XGBoost)neural networks is proposed.Combining the advantages of LSTM and XGBoost,we first use LSTM and XGBoost for flight arrival delay prediction respectively,and then use polynomial regression for weighted combination to obtain the final prediction results.Through comparison experiments on the U.S.public flight dataset,it is shown that the accuracy of the flight delay prediction by this model can be improved.(3)By analysing the analysis of the intrinsic factors affecting flight delays,and mining the propagation impact of flight delays in the airport ground network,then a BiLSTM-Catboost flight delay prediction model(FBLCM model)based on flight delay propagation and attention is proposed.The delay propagation impact is obtained by calculating the impact of arrival delays of preceding flights in the same flight chain on the departure delays of subsequent flights.Adding delay propagation to the model and also adding an attention mechanism to optimise the model feature extraction effect to obtain more accurate flight delay prediction results.Comparative experiments on a publicly available US dataset show that the model has improved the accuracy of the prediction results.(4)Based on the above proposed flight delay prediction models for different types of flight delays,a flight delay prediction system has been designed and implemented on the basis of flight chains.According to the information input by the user,this system calls the database to identify whether there is a flight chain or not,so that different models can be used to predict flight departure delays and arrival delays.Finally,the prediction results and related information are presented to the user.