From the black box to the glass box: Using unsupervised and supervised learning processes to predict user engagement for the airline companies

TLDR

Firms collect large amounts of user‑generated content such as social media posts to analyze consumer opinions, but lack an analytic framework for unstructured UGC. The study applied unsupervised and supervised AI to 680,410 airline‑related tweets and 4,961 retweets, modeling engagement while evaluating word count, time gaps, hashtag usage, and topic extraction. Results showed positive eWOM sentiment for United, Delta, and Alaska Airlines and negative sentiment for Southwest and Hawaiian, offering managers a detailed guide for leveraging unsupervised and supervised analytics to boost engagement.

Abstract

Firms collect an enormous amount of user generated content (UGC), such as social media posts, to analyze consumers' unfiltered opinions regarding brands and firms. A challenge in analyzing unstructured UGC is the lack of analytic frame. By adopting both unsupervised and supervised learning processes for using artificial intelligence (AI), we collected 680,410, tweets related to airline companies (United Airlines, Delta Airlines, Southwest Airlines, Alaska Airlines, and Hawaiian Airlines) and analyzed 4961 retweets to predict user engagement levels on Twitter. Rooted in the electronic word-of-mouth (eWOM) perspective, the results of this study indicated that consumer sentiment was positive for United Airlines, Delta Airlines, and Alaska Airlines, whereas it was negative for Southwest Airlines and Hawaiian Airlines. We also examined the effects of word count, gaps between the tweet generated date and the retweeted date, the number of the hashtag(s), and extracted topics on predicting the level of user engagement. Ultimately, this study provided a detailed guide to mangers on how to use an unstructured data analysis procedure incorporating both supervised and unsupervised learning processes.