Airline Passenger Characterizations Using LDA Topic Modeling

This study investigates airline passenger characteristics with respect to cultural differences, seat class differences, and airline business model differences while applying latent Dirichlet allocation (LDA) topic modeling to 40,885 passenger-authored online reviews. Because passengers experience airline services at service encounters in the whole service sequence as their perceptions from service encounters are directly enclosed in online reviews, the extracted topics can be well represented as the service encounters. The LDA topics become groups of words with similar characteristics and they are statistically distributed by word frequencies from online reviews containing passenger perceptions of airline services. Then we quantitatively compare and analyze airline passenger characteristics based on the importance of service encounters, which is expressed by probabilities of the word frequency distribution. The results show that airline passengers are differentiated by cultural characteristics and airline business model characteristics, but not by seat class characteristics.