Hybrid DEEP-learning model for resilient tourism enterprises

The tourism business is increasingly encountering disruptions due to external shocks, fluctuating demands, and changing tourist sentiments. Whereas conventional risk prediction models are mostly static and less responsive. The study aims to design a real-time decision support tool through a novel CRISP-LSTM11 Cascading Risk-Integrated Strategic Planning using Long Short-Term Memory model, integrating hierarchical LSTM structures with attention-driven multimodal fusion of tourism statistics, online reviews, and GDELT22 Global Database of Events, Language, and Tone events. The results reveal that CRISP-LSTM performs better than conventional models, with better precision, recall, F1-score, and AUC (0.86, 0.90, 0.88, and 0.92, respectively). Simulations on high-risk scenarios indicates reduced economic loss (40%) and recovery time (six months). The real-time CRISP-LSTM significantly enhances strategic results with extreme disruption scenarios. Thus, the study offers a practical and novel approach for tourism enterprise resilience.