Study of Airline Passenger Preference Formation Mechanisms - An Application of a Hybrid Analysis Method Based on Skytrax Ratings and Reviews

This study explores airline passenger preference formation by integrating structured rating data and unstructured review text from over 10,000 Skytrax user comments. A hybrid analytical framework combining binary Logit regression and natural language processing is employed to assess the impact of seven key service dimensions—including value for money, cabin service, and ground service—on passengers’ willingness to recommend. Results show that value for money is the most influential factor, with notable differences in preference across traveler types and cabin classes. Sentiment analysis using a supervised learning model maps review emotions to a -5 to 5 scale, revealing a U-shaped distribution: extremely satisfied or dissatisfied passengers are more likely to leave reviews. Text mining further shows that dissatisfaction often relates to delays and pricing, while satisfaction is tied to crew service and comfort. The study offers actionable insights for airlines to identify service pain points, optimize experience for different passenger segments, and enhance customer satisfaction through data-driven decision-making.