Research On Key Technical Of Air Passenger Travel Based On Machine Learning Method

At present,machine learning and big data technologies have made great progress,and machine learning methods and other related data analysis technologies have been widely used in various fields.Especially in the context of the rapid development of the civil aviation transportation sector,the intrinsic drive of smart airport construction and green travel of air passengers has led to the growth and accumulation of data related to industry production,and the use of machine learning technology has enabled airport operation management and related service quality improvement.However,due to the complexity of air passenger travel-related issues and the intertwined nature of the associated operations,the desired results are often not achieved using some of the traditional methods.This paper focuses on machine learning methods as the key technical implementation route to build a new modeling method for management and decision making,which will provide a better solution to air passenger travel related problems.The seminar focused on a number of issues related to air passenger travel such as passenger throughput forecasting,flight passenger forecasting,travel index forecasting and ground service security staff scheduling.This paper focuses on the use of existing machine learning technology methods to improve and construct new methods based on machine learning technology,and then realize the application of these new methods to solve specific problems related to airport operation management and air passenger travel.Firstly,we sort out and analyze the problems of smart airport,air passenger travel,passenger throughput,airline flight seat rate,and ground staff scheduling,at the same time,we also discuss machine learning methods,Fuzzy theory,information theory,and extreme value theory.Then,based on time series,mutual information,fuzzy,extreme value theory intergrated and improved machine learning method models,proposing prediction methods based on long-short memory and support vector regression,improved K-means clustering algorithm,prediction methods based on mutual information and support vector regression,and improved Bayesian network prediction method based on fuzzy maximum mutual information.Finally,these methods were applied to throughput forecasting,flight passenger volume forecasting,air passenger travel index forecasting and ground service security staff optimization scheduling problems respectively and achieved better results.The main research work of this paper is described as follows:1)Aiming at the air passenger throughput prediction problem,a support vector machine regression prediction method based on the improved long and short-term memory network(LSTM)is proposed based on the characteristics of air passenger throughput prediction time series,which improves the long-term dependency problem existing in RNN,considering the long and short-term memory network(LSTM).LSTM as a dynamic recurrent neural network,it can better respond to the dynamic change of air passenger throughput results,while the SVR has the correction feature of eliminating many redundant samples,which greatly reduces the transmission of prediction errors over time through the integration of the advantages of the two machine learning methods.The experimental analysis shows that the LSTM-SVR model is feasible and has higher prediction accuracy,and the method is an aid to the decision making of airport scale and resource matching construction.In addition,according to the layout of airport network,considering the impact of airport network route throughput,the construction of flight passenger prediction model,providing technical means for the prediction of flight passenger volume in airline operation that can effectively help airlines understand their customers better,and this research result will bring some help to the improvement of airline route operation revenue.2)Aiming at the research of air passenger travel evaluation issues,the main research work is as follows: Firstly,this paper puts forward the concept of air passenger travel index and index level for the first time,they can well reflect the airport busy degree and provide a new reference index for airport operation management and passenger travel.Secondly,for air passenger travel index prediction,the key influencing factors are selected using mutual information values,and a support vector regression MI-SVR model method based on mutual information improvement is constructed.The empirical analysis found that after selecting the feature influence factor by mutual information,the common prediction methods have significantly improved the prediction effect,while the improved support vector regression MI-SVR model method proposed in this paper is relatively better.Finally,the empirical analysis finds that after selecting the characteristic influence factors by mutual information,the prediction effects of common prediction methods are significantly improved,while the improved support vector regression MI-SVR model method proposed in this paper is relatively better.On the one hand,airports can choose to arrange the operation and maintenance of airport-related facilities and equipment or other construction activities during relatively less busy time periods,which can minimize the impact of such work on airport operation and management;on the other hand,for the peak travel time of passengers,the corresponding internal and external resources can be fully deployed.On the other hand,the internal and external resources of the airport can be fully deployed for the peak travel times.The air passenger travel index is not only applicable to small airports,but also to large hub airports,which is a reference value for passengers to choose the right time by air.3)Considering the small probability events affecting air passenger travel index factors,while using fuzzy theory,a Bayesian network air passenger travel index prediction method based on the improved idea of fuzzy maximization of mutual information is proposed.This method is used to improve the class effect of classifier classification by fuzzy construction of virtual sample set i.e.,randomly grouping the data,equalizing the data of occasional phenomena,and letting them sprinkle in the chaotic data set,using the original redundant data of the data to reduce and strengthen the role of relevance data.The empirical analysis finds that the improved Bayesian network model method proposed in this paper is feasible and effective,and the method is not only applicable to accurate prediction with a large amount of historical air passenger travel index data,but also performs better when a small amount of sample data is available.In addition,the use of elbow method to improve the clustering method K-means for air passenger travel index,influence factors clustering analysis,empirical analysis of clustering effect is better.4)Aiming at the ground service staff scheduling problem under the extreme value distribution of air passenger travel index,the main research work is as follows: Firstly,according to the ground service staff scheduling characteristics,combined with extreme value theory and classical intelligent optimization algorithm,the ground service staff scheduling model considering the extreme value distribution of air passenger travel is proposed.Secondly,the intelligent optimization algorithm is used to solve the solution to obtain a better ground service staff scheduling scheme.Finally,the empirical analysis is carried out,and the experimental results show that after obtaining the extreme value distribution by machine learning method for travel index prediction,the use of intelligent optimization algorithm scheduling is more efficient than manual standard scheduling,and the comparison and analysis of whether to consider the extreme value distribution for ground service security personnel intelligent optimization algorithm scheduling,considering the extreme value condition scheduling and according to the mean value condition,which makes the total amount of ground service security work in the established working period is reduced,and the efficiency of personnel utilization is improved.It is explored that an airport ground service security service 53 weeks as a working period,workload(in terms of hours)compared to reduce about 8.1%,with good management benefits.In addition,the questionnaire survey found that the scientific ground service security staff scheduling scheme is conducive to improve employee satisfaction.The research on the application of big data of airport and airline operations based on machine learning methods is an important initiative to promote the construction of intelligent civil aviation business.This study constructs a good new management decision-making model method,improves the efficient operation of the airports and airlines,and meets the demands of passengers’ travel comfort and convenience,which is of great significance in theory and practical application.