Forecast Research On The Demand For Checked Baggage Of Departing Passengers At The Airport Terminal Based On Data Driven

According to the statistics bulletin of civil aviation administration,China airports have checked in over 1 billion passengers,and 292 million baggage[1].Many airport facilities,especially the check-in facilities have been overburdened.Passenger baggage check in is an important part of the check in process,its efficiency directly affects the time passengers stuck at the terminal.In addition,with the improvement of self-service luggage,self-service sorting and other technologies,many new problems remain to be solved.The present study at the airport baggage operations is mainly focused on information technology and advanced equipment,the check-in process simulation and resource optimization.There are few researches focused on luggage check in from the perspective of demand forecasting,and most of the research are based on modeling and simulation,few research results is based on real data.Results obtained through real data analysis is different from results obtained through simulation,which cannot be used in real practice.This research combines mass actual operational data from Kunming Changshui International Airport,analysis the departing passengers baggage check in demand distribution over time,investigate the interaction mechanism between departing passengers and luggage check in.Base on which,we constructed the forecasting model of luggage check in demand over long time and short time.Finally,through empirical analysis,we testified the efficiency and utility of the research result.The results is listed as follow:(1)In order to studies the principle of luggage check in demand,we pre-processed the mass actual operational data from Kunming Changshui International Airport,including data deletion,abnormal data recognition,data interpolation,etc.We calculated the luggage check in demand and departing passengers over different times.Based on this,we investigated deep into the time distribution property of luggage check in demand,and found that luggage check in stream shows similarity in period of day,week,and short terms.Taking advantage of Fisher ordered clustering method,we divided luggage check in stream into 5 state intervals.Meanwhile,we have elaborated the interaction mechanism between departing passengers and luggage check in demand.We have found that departing passengers and luggage check in demand are related in a certain time delay.(2)Based on the self-relevance and periodicity of luggage check in stream,we constructed long-term forecasting model for departing passengers luggage check in demand based on SARMA model to forecast the luggage check in demand in different time of the next day.The experiment result shows that SARMA model is capable of depicting the internal property of random variable.The model is precise and robust.(3)Based on the flight time table and history data of luggage check in demand,we have constructed forecast model of luggage check in based on GBDT ensemble learning model to forecast the luggage check in demand of the next time period.First,we have constructed high dimension feature engineering to acquire effective and sufficient model input variable.Second,according to the different survey methods of luggage check in stream and passenger stream,we have designed multiple GBDT forecasting models.At last,combining the actual operational data of Kunming Changshui International Airport,we have trained,testified and analyzed multiple models.We have found that compared with 10 min and 20min time interval,5 min time interval of passenger stream survey shows best forecasting result.Meanwhile,modeling and processing the luggage stream separately according to their status can decrease the feature variable dimension,effectively extract highly relevant feature variable,and further enhance the forecasting accuracy.Based on the analysis above,we have reached the best model plan,and analyzed the forecasting result of the plan.The experiment result shows that the GBDT short term forecasting model boasts high forecasting accuracy,can satisfy the actual operational demand of the airport.GBDT ensemble learning method is capable of guarantee high forecasting accuracy and robustness,and in the meantime recognize the influence of input variable towards forecast result,revealing the interactions between various factors in complicated problems.With the ability to accurately predict over a long time,the airport can allocate luggage transport resources and make production plan ahead of time.With the ability of accurately predict over a short time,the airport can optimize the departing process dynamically and allocate terminal resources intelligently.The long term and short term forecasting model constructed in this article boasts high accuracy and high utility.It can be used as decision support for constructing intellectual airport.