The road not taken: Representing expert knowledge for route similarities in sustainable tourism using machine learning

Abstract As recreational tourism in rural areas rises in popularity, overtourism, and crowding pose growing challenges, impacting both society and the environment. To support sustainable smart tourism, an information system for visitor management offers a valuable approach. A significant challenge in this context is the identification of suitable alternatives to congested areas. This paper proposes a method to calculate route similarities with distance-based algorithms and machine learning models using descriptive data to redirect visitors to less-crowded paths. A case study in a nature park validates the approach, using real-world hiking data from an online outdoor platform. Expert surveys on route similarity are used to train the models and evaluate the results. Machine learning significantly outperforms traditional similarity algorithms, achieving up to 117% higher R 2 values (0.448 vs. 0.206), 26% lower MSE values (0.530 vs. 0.719), and 40% higher Spearman correlations (0.699 vs. 0.498). The random forest regression model yields the best results. This research provides a foundation for future efforts to enhance sustainable tourism by offering a data-driven approach to identifying alternative routes that align with visitor preferences.