Airline and Railway Customer Lifetime Value Estimation A Federated Learning-based Civil Aviation Passenger Value Analysis Method and MaaS Construction Considerations in the Epidemic Background

Abstract

Data and artificial intelligence technology can help transport operators such as airlines and railways to increase their operating income. Recognizing high-value passengers is particularly important for transportation operators because they can bring more than 50% of profits. However, unpredictable factors such as the global COVID-19 pandemic epidemic and data privacy leaks have challenged the assessment of high-value passengers. The aim of this thesis is to adequately fill the customer value assessment gap under unpredictable transportation factors. The thesis adopts the form of a journal, and it consists of six papers. Papers 1 to 3 provide an in-depth study on how Big Data machine learning technology can help in the evaluation of passenger value. Paper 1 shows the Bayesian belief network of an airline content recommendation model. In this paper, the static causality model is transformed into a dynamic causal model to show its enhancement effect on the prediction of consumer choice preferences. Paper 2 proposes an optimized airline passenger value assessment model. A fit test is conducted with TravelSky’s value score to improve the accuracy of passenger value assessment. Paper 3 adopts a two-stage modelling method. In the first stage, series regression, gradient boosting decision tree and neural network model are used for modelling the passenger value evaluation. These models are integrated to improve the accuracy of passenger value assessment. Papers 4 to 6 cover management theory using federated learning methods. Paper 4 proposes a new classification model based on federated learning, which is used for high-value passenger recognition in the civil aviation industry. This model enhances the accuracy of passenger value assessment. Paper 5 uses federated learning technology to simulate the willingness to pay for upgrade services and based on the results, facilitate the possibility to accurately predict the precise population that is willing to perform aircraft upgrade services, thereby improving the accuracy of customers’ willingness to pay for upgrade services. Finally, Paper 6 proposes a method of applying vertical federated learning technology to help airlines to identify target customers. It also extends the discussion from civil aviation to railway and proposes a concept of constructing the mobility as a service (MaaS) platform for the future.

Date of Award	8 Feb 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Dong Xu (Main Supervisor) & RMS UnKnown (Co Supervisor)