Application of spatio-temporal clustering algorithm based on big data in the analysis of tourists' behavior in the palace museum: a case study of Ctrip and TripAdvisor reviews

Abstract

A grid-based spatio-temporal clustering framework is developed to extract historical visitor behavior patterns from large scale Ctrip and TripAdvisor datasets focused on the Forbidden City. The methodology incorporates trajectory preprocessing using adaptive filtering and segmentation, followed by density-based spatial clustering on a uniform grid. The spatio-temporal adjacency matrix and clustering similarity measure are used to determine the transient and persistent tourist hotspots. By using the periodicity index and autocorrelation function, the temporal analysis can reveal the daily and weekly cycles of tourist flows. Validated by the silhouette index and the Davies-Bouldin index, the framework demonstrates the ability to partition core attraction areas, reconstruct major tourist paths, and quantify path popularity. The results highlighted the spatial heterogeneity of the site and showed high-traffic visit intervals on weekends and holidays. The method can accurately identify congested and underutilized areas, thus providing data for heritage tourism resource allocation and adaptive management. These results show that advanced computational mining of heterogeneous trajectory data can greatly improve the interpretability and resolution of behavioral analysis of complex cultural destinations.