Optimizing airline service performance: Predictive modeling of passenger satisfaction via binary logistic regression

Abstract

This study investigates the determinants of airline passenger satisfaction using a large-scale dataset (N = 25,976) sourced from Kaggle, applying binary logistic regression to assess the influence of sociodemographic characteristics and service-related variables. Descriptive statistics reveal a predominantly loyal, business-oriented clientele, with a slight female majority and a preference for business and economy cabin classes. Regression results show that 18 of 23 predictors significantly influenced satisfaction at the p < .05 level. Notably, passenger type of travel (OR = 16.298, p < .001), customer loyalty (OR = 7.738, p < .001), and online boarding (OR = 0.552, p < .001) emerged as the most influential determinants. Digital conveniences (e.g., online booking, Wi-Fi access) and operational aspects (e.g., check-in service, legroom, baggage handling) significantly shaped satisfaction more than traditional physical comfort. The logistic regression model achieved an accuracy of 87.1%, specificity of 83.4%, sensitivity of 90.0%, and AUC of 0.926, demonstrating high predictive validity. These findings suggest that airlines must prioritize seamless digital experiences and consistent service delivery to retain passenger satisfaction and loyalty in an increasingly competitive market.