Spatial And Temporal Distribution Prediction For Airport Passenger Flow

In the past decades,the economy developed so rapidly that people are having increasingly quality of life,more and more people choose airline as their primary transportation,this lead a huge pressure to the airport.For the more efficient utilization of the resources in the airport,the directors expect using the big data technology,coupled with machine learning,to improve the efficiency of production.Airport has huge passenger flow each day,the movement appears to be dynamic and have an uneven distribution of time and space,bringing the huge service burden to the airport.The security check,emergency response,check-in,baggage tracking want to predict the passenger flow distribution in advance,in order to schedule the manpower and material resources for better user experience.Based on the real situation,the article analyse and visualize the real data,extract the main feature that influence the passenger flow distribution,including history features,flight departure features and wifi ap location features.Using the history features as the basic model,add other features to develop the prediction model.Dividing the single regression model into five regression models by location features to decline the feature dimensions,avoid overfitting and raise the accuracy of the prediction.Using linear regression,gradient boosting regression tree and recurrent neural networks to predict the future distribution on the real data.Analyse the prediction result and compare strengthens and weakness of the different methods.