An Airline Passenger Satisfaction Prediction by Genetic-Algorithm-Based Hybrid AutoEncoder and Machine Learning Models

The evolving landscape of the Business-to-Client (B2C) model across the globe is reshaping service delivery paradigms and transforming consumer perceptions of service providers, thereby revolutionizing customer experiences. This paradigm shift directly impacts airline companies that offer multiple service tiers, necessitating ongoing promotional strategies to attract and retain customers. Moreover, in addition to attracting new passengers, it is equally vital for airlines to retain existing ones. Therefore, comprehensive research is imperative to comprehend customers’ perceptions and conduct post-flight customer satisfaction surveys to delve into the factors influencing their decision-making processes. By gaining insights into these crucial causal factors, airlines can tailor their services to better meet customer expectations and enhance overall satisfaction levels. To address these challenges, this paper proposes a hybrid model comprising Deep Autoencoder (DAE) and Genetic Algorithm (GA) techniques for optimizing feature extraction. Utilizing eleven Machine Learning (ML) models as baseline predictors, the study endeavors to forecast passenger satisfaction levels. Furthermore, each ML model is intricately combined with the AE-GA optimization framework to conduct in-depth customer satisfaction experiments. Conducting a 5-fold cross-validation analysis in each experimental setup, the study highlights the efficacy of the proposed optimization strategy in significantly enhancing the predictive performance of ML methods in forecasting customer satisfaction levels.