Research On Flight Delay Prediction Based On EMD And SVM

At present,flight delays seriously restrict the sustainable and healthy development of China's civil aviation transportation industry.In particular,the frequent occurrence of sudden large-area flight delays has brought serious consequences.Therefore,it is very necessary to study and analyze flight delays and achieve effective prediction,so as to ensure the development of the civil aviation industry.In view of the current situation that domestic airport flight delays are difficult to accurately predict,based on the delay data of inbound and outbound flights at hub airports,proceeding from the perspective of historical data,we conduct research on the law of delay to improve the accuracy of flight delay prediction.First,the data is preprocessed through standardized statistics and converted into flight delay sequences after processing.The EMD algorithm is used to analyze the characteristics of hub airport delay laws.According to the research and analysis,the ICEEMDAN algorithm with the aid of noise-assisted analysis is selected to make up for the defects of the EMD algorithm and improve the ability to process and decompose the delayed sequence.However,the decomposition results found that the modal component has an amount of noise that is difficult to eliminate.For this,an improved ICEEMDAN denoising algorithm is proposed.The correlation coefficient analysis is used to determine the correlation between the modal component and the original sequence and the noise distribution in the component.Residual problems.The results show that the proposed improved algorithm implements the filtering and decomposition of delay sequences and obtains the regular characteristics of flight delays.Next,the support vector machine regression multi-step prediction model is established for the modal components obtained from the processing.Under the premise of uniformly selecting the radial basis kernel function of the model,the problem of model decision complexity caused by too many kernel function types is solved.At the same time,the predictive ability requires the corresponding parameters of each component model as the support.Based on a hybrid algorithm composed of grid search and comprehensive particle swarm algorithm,the idea of cross-validation is introduced to optimize the model parameters according to the characteristics of each component.The verification and analysis show that the training results of each component model are good,which is conducive to the multi-step prediction of flight delays of the combined model.Aiming at the established combination model,the forecasting ability decreased in the severe fluctuation range of the delay sequence.After statistical analysis,it is found that the sequence fluctuation interval has a strong correlation with the corresponding severe meteorological conditions,and a flight delay prediction model combining gray correlation analysis is proposed.This model analyzes and judges the factors of severe weather conditions by grey correlation analysis method,and assigns the weights of each factor to realize the level prediction of flight delay under severe weather conditions.It is conducive to repairing the problem of insufficient flight forecasting ability of the combined model under severe weather conditions and improving the accuracy of flight delay forecasting.