How Social Media Information Characteristics Shape Destination Choice Intention: An LDA–SEM–XGBoost Framework

The widespread use of social media has reshaped the production, dissemination, and consumption of tourism information. Tourists’ destination selection is jointly influenced by platform content characteristics, algorithmic recommendation, emotional expression, and information credibility. This study constructs a three-stage analytical framework of “information feature quantification -behavior mechanism modeling-destination prediction feedback”. LDA topic modeling and VADER sentiment analysis are used to extract tourism content features from Weibo, Douyin, and Xiaohongshu, while a structural equation model is applied to examine the paths and influence intensities of information credibility, communication breadth, and emotional tendency on destination selection intention. An XGBoost classifier integrating multi-source social media features is then employed as a diagnostic tool to rank feature importance and assess the relative contribution of information characteristics to destination discrimination within the sampled set. The results show that emotional tendency and information credibility have the strongest positive driving effects, and multi-feature fusion improves prediction accuracy. Because social media diffusion depends on mobile wireless networks, antenna-supported terminals, and electromagnetic information propagation, communication speed and platform reach are interpreted from an engineering-information perspective. The findings provide methodological support for tourism marketing prioritization and feature-driven strategy design.