Predicting Tourist Satisfaction in Film Tourism: A Machine Learning Approach to Sustainable Destination Management

The driving mechanism of tourist satisfaction in film tourism is complex, and traditional statistical methods have limitations in in-depth analysis of its nonlinear relationships.To break through this methodological bottleneck, this study aims to construct an analytical paradigm with high prediction accuracy and in-depth explanatory ability.Based on 400 valid questionnaires collected at Hengdian World Studios in China, this study innovatively constructs a Gradient Boosting Decision Tree (GBDT) prediction model and combines it with the cutting-edge SHapley Additive exPlanations (SHAP) interpretability framework.The results show that this integrated model exhibits excellent predictive validity, achieving an Area Under the Curve (AUC) value of 0.887 in 10-fold cross-validation.It has improved by 7.5% compared with the logistic regression model used as the baseline (main results).More importantly, the SHAP analysis identifies "service staff quality" as the primary determinant of satisfaction; it also reveals that the second most important factor-"scenic environment attractiveness"-has a significant nonlinear "threshold effect": its contribution increases sharply after the score exceeds the medium level (key insight).This study provides a new, high-precision, and interpretable analytical tool for tourist behavior research.Meanwhile, it offers specific and operable data-driven insights for film tourism destinations on how to conduct accurate service optimization and resource allocation to promote sustainable development.