Short-term Passenger Flow Forecasting For The Check-in Process In The Airport Based On Time Series

In the transition process of the airport to “smart” and “digital” operations,in order to improve the service quality in the terminal building and achieve efficient business operations and rational allocation of resources in the terminal building,the airport must accurately estimate the number of passengers in the terminal building.Only in this way can the service resources in the terminal be reasonably arranged during the departure of the passengers,solve the passenger's dilemma in time,reduce the congestion during the peak period,reduce the waiting time of the passengers,and fundamentally improve the passengers' travel quality.Therefore,relatively accurate forecasting of short-term check-in passenger traffic in the airport terminal is an important basic guarantee for improving service quality,enhancing business efficiency and rationally allocating resources.From the multiple production systems of the airport,obtains operational data related to flight,passengers,check-in,security and other aspects related to the research objectives.Through data flow familiar with data,data analysis is performed on relevant factors that affect the value of passenger traffic.Based on the analysis of the main influencing factors of passenger flow change,the short-time hourly passenger flow is taken as the research object,and the classification and co-integration theory are used to construct the time series and dependency analysis based on the DOW characteristics.The ARIMAX model with dynamic regression of input variables is used to predict the number of passengers on duty in the terminal for short periods of time.The experimental results show that the prediction results of this model have less information and higher fitting than the results of a single time series prediction model,which improves the prediction accuracy and reduces the prediction bias.In the process of analyzing the potential variation of passenger flow,it is found that the law of time series shows a high degree of feature similarity,such as nonlinearity,periodicity and uncertainty.Therefore,establishes a short-term value-based passenger traffic forecasting model based on wavelet analysis for LSTM networks.The model uses the hourly passenger flow as the forecasting unit,uses the wavelet analysis technology to transform and reconstruct the time series,and uses the deep LSTM network model to train the reconstructed time series.Finally,each predicted value is superimposed to obtain an average value,that is,an hourly passenger traffic prediction value is obtained.The experimental results show that compared with the ARIMAX model,the prediction accuracy of the model is higher and the fitting effect is better.Both groups of models effectively predict the number of check-in passengers in the terminal building for a short period of time,which provides indispensable decision support for the dynamic allocation and optimization of resources in the terminal building,and has certain guiding significance for reality.