Adaptive resilience in crisis: A neuro-economic deep learning framework for dynamic customer loyalty management in the airline industry

Airline markets are increasingly exposed to crises of varying frequency, heterogeneity, and severity, necessitating adaptive, data-driven strategies for managing Large-Scale Crisis Adaptive Capacity (LCAC), particularly in relation to customer loyalty. This study proposes a neuro-economic–informed deep learning framework that integrates sequential behavioral modeling with adaptive decision optimization to predict and manage loyalty under crisis conditions. Using 129,880 customer records spanning stable and crisis-affected periods, four algorithms were evaluated: Long Short-Term Memory with Reinforcement Learning (LSTM-RL), Long Short-Term Memory (LSTM), K-Nearest Neighbors (KNN), and Naïve Bayes (NB). To capture emotional volatility, a crisis sentiment layer was incorporated, derived from customer-generated textual data and treated as a latent behavioral proxy. Sentiment was directly observed for 6.7% of customers and imputed for the remainder using similarity-based methods. Results show that the hybrid LSTM-RL model achieved a test accuracy of 0.91, outperforming traditional models (0.84–0.86). Reinforcement learning enabled dynamic adjustment of retention actions in simulated crises, yielding up to 28.4% improvement relative to a no-intervention baseline. The study further introduces the Crisis Amplification Index (CAI), which quantifies how service attributes differentially influence loyalty during crisis versus stable periods, and explores a blockchain-enabled loyalty mechanism for transparent incentive allocation. By integrating economic crisis theory, behavioral economics, and neuro-economic concepts within an adaptive machine learning architecture, this research provides a methodologically integrative—though not neurobiologically direct—approach to airline loyalty management under crisis-induced uncertainty.