A Hybrid AI-driven Text Analytics and DEA Approach to Airline Efficiency Evaluation Based in Online Reviews

Abstract

The airline industry plays an important role in enhancing global connectivity and driving economic growth. However, this sector records low customer satisfaction rates compared to other service industries. This issue highlights the need for operations and service quality improvement based on customer reviews in the airline industry. This research uses online reviews of customers and employees to evaluate the efficiency of airlines, contrasting with existing methods that focus on structured data for this purpose. The study uses BERTopic, a Generative AI tool, for topic modeling on review datasets to highlight critical aspects of service quality. Then, the proposed algorithms categorize each review statement under identified topics and assign sentiment polarities on a five-point scale, later converted into fuzzy variables. These variables serve as parameters in the Slack-based measurement of the Fuzzy Network Data Envelopment Analysis model, which evaluates airline efficiency across three divisions: employee, customer, and sales. The research integrates insights from unstructured data into the airline efficiency evaluation model, offering a comprehensive assessment beyond traditional financial and operational metrics. Implemented on large review datasets about major U.S. airlines, the evaluations highlight the higher accuracy of the presented hybrid AI-driven text analytics and Network DEA methodology compared to statistical text analytics methods. The findings demonstrate the utility of unstructured data in identifying strengths and weaknesses of internal processes and services, as well as providing a foundation for targeted improvements in the airline industry.