Quantifying tourist perception of cultural landscapes in traditional towns in China using multimodal machine learning

作者：Xiaojuan Zheng, Yan Huang, Zhiming Xie, Aimin Zheng · 发表于：npj Heritage Science · 年份：2025 · DOI：10.1038/s40494-025-01907-4 · 被引用次数：6 · 研究领域：Advanced Technologies in Various Fields、Diverse Aspects of Tourism Research

Cultural landscapes shape tourist perceptions. Quantitative research on different types of traditional towns is still limited. This study uses multimodal social media data to explore this issue. It employs XGBoost and SHAP to analyze textual, visual, and rating data from 114 traditional towns. The study quantifies how environment, infrastructure, and experience characteristics influence tourist ratings and sentiment values. The results show that integrating text and images improves feature representation. Tourists focus on different aspects depending on the type of traditional town. Waterfront traditional towns emphasize water features. Ecological traditional towns prioritize the natural environment. Commercial traditional towns emphasize the consumption experience. Religious cultural traditional towns emphasize the cultural atmosphere. Among all features, comfort and aesthetics have the strongest impact on tourist perceptions. This study demonstrates the value of multimodal data and interpretable machine learning in cultural landscape research. It also provides data support for cultural heritage conservation and tourism management.