Research On Flight Volume Prediction Based On ML-SD

The"14th Five-Year Plan"period put forward higher requirements for aviation power,and defined the new direction of the development of civil aviation transport,including civil aviation safety,infrastructure construction and other six areas.Now that the era of COVID-19is over,international air routes have resumed and the number of flights has basically returned to the level of 2019 since China opened its borders on January 3,2023.Flight volume is an important index to measure flight volume growth and airport operation.Scientific flight volume forecast can provide direction for future development decision-making of civil aviation industry,and provide important basis for industry foundation planning and construction.In addition,the short-term forecast of flight volume can effectively assist the operation support departments of airports,air traffic management and airlines to carry out fine management and develop dynamic adjustment strategies in advance;Long-term forecasting can provide a reasonable plan in advance for the need for new or expanded airports.This text first analyzes the factors affecting the flight volume,and finds that seasonality has a great influence on the monthly flight volume.The linear and nonlinear characteristics shared by the monthly flight data of"one city,two airports"and the renovated and expanded airports are analyzed by the unvariable linear regression fitting method,and the airports that are close to saturation need to be built or rebuilt or expanded as an example.For two characteristics,two non-homologous SARIMA and BP neural network models were adopted to consider the limitations of a single prediction model.The combined forecast model of SARIMA-BP neural network was constructed to forecast the flight volume respectively.After the feature identification of time series data,after difference and ADF unit root test,SPSS technology is used to quantitatively analyze the long-term,seasonal,volatility and irregular change trend of flight volume.After two seasonal differences,The order of the linear model is SARIMA（1,1,2）（3,2,2）₁₂with the aim of minimizing BIC.The nonlinear three-layer BP neural network prediction model was constructed,and the nonlinear tansig function was selected as the transfer function to prevent the overfitting of the model and the addition of Bayesian regularization.Through normalization,model training and testing,the inverse normalization was performed to output the prediction results.In order to further improve the prediction accuracy,the two models were combined in series and parallel,and the model with higher prediction accuracy was selected to complete the short-term forecast of flight volume.The accuracy test of the forecast model was evaluated by the MAPE index of the forecast flight volume in 2019.The prediction results showed that the MAPE of the single prediction model SARIMA model was 2.23%,the MAPE of the BP neural network model was 3.47%,and the MAPE of the combined model was 1.74%with the higher"series"accuracy.After learning the rule of monthly flight volume from 2008 to 2019,the above models skip the years from 2020 to 2022 under the influence of COVID-19,and take the flight volume in January 2023 as the known quantity to predict the flight volume in the next 11 months,realizing the forecast of the monthly flight volume in 2023.It provides some reference value for airport fine management and human resource arrangement.In order to explore the influence of external factors on flight volume and realize long-term flight volume prediction,this paper combined the influence of machine learning predicted value,GDP,population,travel decision and other external factors on flight volume,built a system dynamics model to achieve multivariate integration,and predicted the flight volume of Chongqing Jiangbei Airport in the next ten years.With reference to the current three-runway operation mode of Chongqing Jiangbei Airport and the peak hour capacity approved by the Civil Aviation Administration,the maximum annual takeoff and landing capacity is 373,000.Assuming that no new airport is built or expanded,the total number of flights is expected to exceed 373,000 in 2027.Considering seasonal factors,the maximum capacity of the airport will be exceeded for the first time in January 2025.To provide auxiliary decision-making for Chongqing Airport Group in planning and development of Chongqing Jiangbei Airport.