In this study, airline passenger satisfaction was predicted using the Random Forest technique. For this purpose, an open-access dataset consisting of 129,880 passenger observations was used. The dataset includes demographic characteristics, travel information, operational indicators, and evaluations of perceived service quality. Passenger satisfaction was treated as a binary outcome and was estimated using a tree-based classification framework. Model performance was evaluated using accuracy, precision, recall, F1 score, and threshold-independent metrics including ROC–AUC and PR–AUC. The results were analyzed comparatively with a logistic regression baseline model, and a 5-fold cross-validation procedure was applied to assess predictive robustness. The Random Forest model demonstrated high discriminative performance (Accuracy = 0.9585; F1 = 0.9618; ROC–AUC = 0.9936) and consistently outperformed the linear reference model. Feature importance analysis, supported by permutation-based robustness checks, shows that passenger satisfaction is primarily shaped by experiential service attributes and digitally mediated service elements. In particular, seat comfort and online boarding emerged as dominant predictors, while demographic and operational variables exhibited relatively lower predictive influence. By combining traditional hypothesis testing with predictive modelling, the study shows that airline passenger satisfaction does not follow simple linear patterns but is shaped by complex interactions among experiential service factors. The findings provide methodological refinement for academic research in aviation and practical implications for data-driven decision-making in airline management.
DOĞAN et al. (Sat,) studied this question.