Model Predicting Airline Passengers Purchasing Seat Upgrade Ancillaries Products on Indonesian Domestic Routes with Random Forest Machine Learning

To increase profitability, airlines are starting to focus on generating revenue not only from ticket sales, but also from the sale of ancillary products. The research focuses only on ancillary seat upgrade products and aims to create a Machine Learning (ML) model to predict economy class passengers who are likely to accept a seat upgrade offer to business class based on historical data on seat upgrade offers received by airline passengers in the 2023 period. The challenge faced in developing a Machine Learning Prediction model is that the number of passengers who take the upgrade offer and those who do not take it is very unbalanced. In previous literature, most investigations covered European, United Kingdom, China routes, most of which use hub and spoke route networks, while in this research try to fill this gap, especially for airlines operating in Indonesia that have domestic routes and use a point to point route network model. The background of domestic routes is the object of research because 80% of the routes flown are domestic routes between cities in Indonesia, and the top 10 routes that contribute the most to ancillary seat upgrade sales are also domestic routes. To build the model, the author uses machine learning for classification, which is combined with Feature Engineering, Data Balancing and Data Dimension Reduction Techniques. Based on an empirical analysis of a historical dataset of passengers who upgraded their seats, it was found that the best combination was provided by Random Forest, PCA, and Stratified Sampling to predict Indonesian passengers who would receive a seat upgrade with 95.6% accuracy.