Research On The Prediction Of Individual Passenger Value In Civil Aviation

In recent years,the number of passengers in civil aviation of China has shown an increasing tread year by year.It has formed a huge customer group.In the face of fierce market competition,the timely discovery of the value of passengers has gradually become the focus of many civil aviation companies.If the passenger's historical consumer behavior information can determine its future market performance and value,it will help airlines to provide passengers with accurate marketing and caring services to enhance passenger satisfaction and loyalty.Firstly,correctly measuring a passenger's market value in the field of civil aviation is the basis of the entire work of this research.Through a more in-depth study of the various measurement method of customer value,this study proposes a parametric model RFUM(Recency,Frequency,Unit-revenue,Mile)that can be applied to evaluate passengers' value in civil aviation.Firstly,the model starts from different dimensions and make a comprehensive measurement of the value of passengers in the civil aviation market.Secondly,through the parameterization process,the different weights are assigned to attributes respectively.Converting the passenger's value into a one dimensional numerical scalar between 0 and 1 provides a certain degree of traversal for civil aviation market decision makers,which can intuitively reflect the difference of different passenger values.Secondly,the prediction problem of customer value in the civil aviation is the core of this research.This thesis proposes a Time aware Multi-task Value Prediction(TMVP)to predict the value of passengers.First,passengers' consumption behavior and consumption value are time-series.The TMVP model attempts to capture such characteristics in a time-aware manner.Secondly,the value of passengers is indirectly reflected by the consumer's consumption behavior.The TMVP model also attempts to automatically capture the indirect mapping relationship between passenger consumption behavior and value through deep learning.Finally,there is a potential connection between the value of passengers and the willingness of passengers to consume.The TMVP model describes the connection between the two by establishing multi-task learning,further improving the accuracy of the moders prediction of passenger value.Finally,the thesis conducted experiments on the actual dataset of an airline.Firstly,in the measurement of passenger value,experiments show that the passenger parameterized value model RFUM proposed in this thesis can comprehensively measure the market value of passengers.Secondly,in terms of passenger value prediction,compared with the traditional predicting model of customer value,the value prediction model proposed in this thesis can also predict passenger value more accurately.