A two stage machine learning approach for Modeling Customer Lifetime Value in the Chinese Airline Industry

This study proposed a two-stage machine learning approach for modeling customer lifetime value (CLV) by integrating the advantages of three traditional machine learning methods (Stage 1) and an innovative method of evidence reasoning (ER) modeling (Stage 2). Unlike previous studies of developing a single CLV model, this study adopted the ER approach (Stage 2) to fuse the results obtained from the Stage 1 (i.e., logistic regression, gradient boosting decision tree, and neural network models) to improve the prediction accuracy. We illustrated the proposed two-stage approach using a case study of Chinese airline passengers. We analyzed a dataset of over 100,000 Chinese airline passengers, which includes 327 variables related to passengers’ travel experience with major airline companies and online travel agents in China. The case study showed that the proposed two-stage approach performed better in comparison with other machine learning algorithms. This study provided airline companies a useful data-driven method to identify high-value passengers and allowed airlines to allocate the CRM resources more effectively.