Revolution trend investigation of tourism destination image with machine learning.

This article focuses on analyzing the tourism destination image of the Dunhuang Mogao Grottoes, a UNESCO World Cultural Heritage site in China, using Latent Dirichlet Allocation (LDA), a topic modeling technique, applied to 16,859 tourists' online reviews collected from major Chinese travel social networks. The study identifies five key image dimensions—Prominence, Uniqueness, Rich culture, Convenience, and Enormous influence—and reveals distinct cooccurrence relationships among keywords within each dimension. Additionally, it examines the evolution trends of these dimensions over 13 years, finding varied patterns including a declining trend in Prominence and ascending trends in Uniqueness and Enormous influence. The research contributes methodologically by integrating topic modeling with text visualization and trend analysis, offering practical insights for destination marketing organizations to tailor image management and monitor changes over time.