What the study found
An event-aware machine learning model using large language models (LLMs, systems that analyze text) and social media data was able to predict cross-city visitor flows. In the Hong Kong case study, the model achieved a testing R-squared of over 85% for daily flows of Chinese Mainland visitors.
Why the authors say this matters
The authors conclude that event-aware visitor flow prediction can help coordinate inter-agency measures and support specialized transport policies. They say such analysis can inform responsive border control operations, dedicated shuttle services to event venues, and on-site traffic management strategies.
What the researchers tested
The researchers proposed a framework to analyze visitor flows under multiple concurrent public events. They used multi-source online information and social media content to extract event features with LLMs, designed social media popularity metrics to capture online promotion and word-of-mouth, and then applied an event-aware machine learning model to predict visitor flows for upcoming events.
What worked and what didn't
The framework was tested with real-world events, social media, and visitor arrival data from Hong Kong. Promotional popularity and word-of-mouth popularity were associated with increased visitor flows, but the specific effects varied by event type; the association was more pronounced for visitors arriving by metro and high-speed rail and less effect was seen for air travelers.
What to keep in mind
The abstract describes a case study based on Hong Kong data and daily flows of Chinese Mainland visitors, so the scope is limited to that setting in the available summary. It also does not provide detailed model limitations beyond noting that the study focuses on multiple concurrent events.
Key points
- The study reports that an event-aware machine learning model could predict daily visitor flows with a testing R-squared above 85%.
- The framework used large language models to extract event features from online information and social media content.
- Promotional popularity and word-of-mouth popularity were both associated with increased visitor flows.
- The effects varied by event type and were stronger for metro and high-speed rail arrivals than for air travelers.
- The authors say the analysis may support coordinated transport and border management responses.
Disclosure
- Research title:
- Event-aware model predicts cross-city visitor flows
- Authors:
- Xiaohan Wang, Zhan Zhao, Ruiyu Wang, Yang Xu
- Institutions:
- University of Hong Kong, Hong Kong Polytechnic University
- Publication date:
- 2026-04-20
- OpenAlex record:
- View
Get the weekly research newsletter
Stay current with peer-reviewed research without reading academic papers — one filtered digest, every Friday.