Interpretable artificial intelligence for travel small and medium-sized enterprises: A threshold-based framework for personalized tourism in Thailand

• Threshold-based AI enables personalization for data-scarce tourism SMEs. • Lightweight framework supports real-time use without cloud infrastructure. • Models achieve over 99 % accuracy with interpretable decision-making. • Human-in-the-loop design enhances trust and operational flexibility. • Framework aligns with SDGs and promotes inclusive digital transformation. Personalized travel demands AI solutions suited for tourism small and medium-sized enterprises in developing countries. This study proposes a threshold-based AI framework that addresses data scarcity, low technical readiness, and infrastructure gaps in Thailand. The model integrates class distributions, nearest neighbors, and neural networks with categorical encoding (−1, 0, +1), reaching 99 % accuracy while preserving efficiency. Human-adjustable thresholds support transparency and interpretability, enabling real-time use without cloud reliance. Empirical results confirm strong calibration and model robustness in low-data environments. The system boosts the competitiveness of small and medium-sized enterprises through direct personalization and aligns with Sustainable Development Goals 8, 9, 10, 12, and 17. Designed for human-centered application, this framework offers a scalable solution for service sectors operating under similar limitations. The approach introduces a practical AI paradigm that balances performance, interpretability, and accessibility—advancing responsible innovation in emerging economies.