Using ensemble classification algorithms to predict airline customer satisfaction

Abstract

The COVID-19 pandemic significantly impacted the airline sector, which has seen a shift in passenger behaviours and a decline in revenue. To navigate this challenging environment and regain customer trust, airlines must prioritise actions that focus on improving customer satisfaction, as satisfaction is a key driver of post-pandemic revenue growth. This paper proposes a novel predictive model for customer satisfaction using ensemble learning techniques. The analysis provides a comparison between the results of single supervised machine-learning methods, such as K-nearest neighbours and decision trees, and those of ensemble methods. The AdaBoost method with a decision tree as the base learner is found to achieve the highest accuracy, at 90.74 per cent. By enabling airlines to proactively address customer concerns and personalise offerings, this model has the potential to significantly improve customer satisfaction and ultimately drive sustainable revenue growth in the post-pandemic era.