The Causal Relationship Between In-Flight Service Features and Air Transportation Passenger Reviews: Interpretable Machine Learning Models

Abstract

The rapid expansion of the airline industry has intensified in-flight services to sustain passenger satisfaction and operational efficiency. This study investigates the causal relationship between in-flight service parameters and total passenger reviews by employing multiple machine learning models; Support Vector Machines (SVM), Neural Networks, Linear Regression, Gaussian Process Regression, Kernel Approximation Kernel, Classification Tree, and Ensemble Learning. In analysis, the feature importances of the boarding/deplaning efficiency, flight crew performance, in-flight entertainment, and Wi-Fi availability considered as key features of in-flight service that might affect the average passenger ratings were evaluated by Shapley and Lime approaches. According to the analysis result, SVM and Linear Regression models exhibit superior predictive performance, evidenced by the lowest Mean Squared Error (MSE) and Root Mean Square Error (RMSE) on test data. Specifically, SVM achieved an RMSE of 0.0385 and MSE of 0.00148, while Linear Regression model followed closely with an RMSE of 0.0368 and MSE of 0.00136. From the analysis results, Shapley and LIME feature important scores obtained from SVM and Linear Regression models highlight the critical impact of boarding/deplaning processes and crew efficiency on passenger satisfaction. From a managerial perspective, the study provides actionable insights for optimizing in-flight resource allocation and improving service delivery. © 2025 The Aeronautical and Astronautical Society of the Republic of China. All rights reserved.