Walking Tourism in Destination Management: Analysis and Prediction of Tourist Preferences Using an Integrated Machine Learning Model

Walking tourism is an important form of thematic and sustainable tourism, especially in rural and naturally attractive destinations. It contributes to diversifying the tourist environments and improving destination management. This paper analyses the role of walking tourism in destination management and uses an integrated machine-learning model to predict tourist preferences. A key focus of this study is identifying the key factors influencing walking tourism preferences, including demographic, socioeconomic, behavioural, and activity-related variables. The methodology of this study is based on an integrated Machine Learning (ML) approach. CatBoostClassifier was used as the primary predictive model, and hyperparameter optimization was performed using Particle Swarm Optimization (PSO). Model interpretability was ensured through SHapley Additive exPlanations (SHAP) analysis, supported by CatBoost feature importance evaluation. This combination enables both high prediction accuracy and transparent explanation of variable influence. The research is based on 467 responses collected through an anonymous online survey. Results confirm that walking tourism is predominantly linked to natural and mountain experiences, which have strong implications for destination planning and tourism product development. The proposed model provides reliable predictions of tourist preferences under class imbalance conditions, achieving a macro-F1 score of 0.5114. Additionally, key factors influencing the choice of walking tours were identified, supporting destination managers in tourist segmentation, tourism product development, and sustainable allocation of destination resources.