Potential High Value Passenger Discovery Based On Booking Behavior Analysis

In recent years,the tide of economic globalization has swept all industries.Airlines are facing increasingly fierce market competition while welcoming development opportunities.Under the pressure of market competition,how to mining and forecast potential high value passenger and turn them into high value trend before competitors has become the main concern of airlines.Nowadays,the civil aviation system has accumulated a large number of passenger booking records.By organizing and utilizing these data through information management and finding out the regular of civil aviation passenger's booking behavior,it is of great significance for airlines to predicate the passenger's value categories,improve core competitiveness and maximize revenue.Aiming at the difficult of the civil aviation passenger's value categories definition,proposed a classification method of civil aviation passengers' value based on clustering,which is based on Passenger Name Record(PNR)data set and Customer Lifetime Value(CLV)theory.At the same time,in order to improve the stability of clustering results,K-means clustering algorithm is improved by repeatedly comparing the distance between clusters and within clusters aimed to adjust the initial clustering centers dynamically,which enhances the sample representability.The experimental results show that the classification method of civil aviation passenger's value based on clustering can mine potential high value civil aviation passenger reasonably,and the improved K-means clustering algorithm has a better clustering effect and higher execution efficiency.Aiming at the problem of low prediction accuracy of traditional classification models due to the behavior similarity between potential high value passenger and low value passenger,Restricted Boltzmann Machine(RBM)is used to extract passenger behavior features and Back Propagation Neural Network(BPNN)is used to learn passenger features and realize the prediction of civil aviation potential high value passenger.The experimental results show that the proposed combined prediction model has a higher prediction accuracy for potential high value passenger compared with traditional prediction model.Aiming at the problem of poor stability of combined prediction model,the complexity of the network is reduced,and the Genetic Algorithm-Simulated Annealing(GASA)is used to find the optimal initial weights and biases of the network accurately.The experimental results show that the improved combined prediction model can predict potential high value passengers more stably and accurately,and has a high practical value that can be used in behavior analysis and prediction problems in different fields.