Abstract
While most human mobility research focuses on the predictable "home-to-work" routine, this study investigates the "irregular" movement of travelers and tourists. By applying information-theoretic measures to a decade of Australian travel surveys, the researchers demonstrate that the predictability of our journeys is deeply tied to our origin, our destination's population density, and—most importantly—our reason for traveling.
Introduction
Most studies on human mobility rely on "daily life logs" (like GPS or call records), which blend regular routines with occasional irregular trips. This makes it difficult to understand the underlying dynamics of non-routine travel, such as tourism or business trips. Understanding these irregular patterns is vital for urban planning, transportation management, and especially infectious disease control, where the movement of "outsiders" into a population can drastically change transmission dynamics. The authors seek to fill this gap by analyzing "meta-populations" (groups categorized by origin) and the specific context (purpose) of their travel to see how predictable these movements actually are.
Datasets and Methodology
The study utilizes two massive, long-term survey datasets from Australia:
- International Visitor Survey (IVS): Data from 2005–2015, covering approximately 40,000 international travelers per year.
- National Visitor Survey (NVS): Data from 1998–2015, covering approximately 33,000 domestic Australian travelers per year.
- Twitter Data: Used as a baseline for comparison, consisting of ~79,000 accounts with precise geo-locations within Australia during 2015.
The researchers used Entropy—a measure of randomness or uncertainty—to quantify predictability. They employed three specific metrics: Information of Visits ($S^{rand}$) based on location counts, Uncertainty of Visits ($S^{unc}$) based on visit probability, and Uncertainty of Displacement Distance ($S^{weight}$) which factors in the distance between locations.
Key Insights: Context is King
Predictability isn't a fixed trait of a person; it changes based on why they are moving. For example, business travel is consistently the most predictable across all regions because it is tied to specific economic hubs.
Other key findings include:
- The "Holiday" Divergence: Europeans visiting Australia for holidays have very unpredictable patterns compared to when they visit for education. Conversely, East Asian visitors are more predictable on holiday than for education.
- Socio-Economic Influence: Domestic travelers from highly populated states (like New South Wales or Victoria) have the most unpredictable patterns, likely due to having more infrastructure and travel options available to them.
- Distance Matters: By including displacement distance in entropy calculations, the authors found that travelers often show a "certainty" regarding distance—tending to favor specific clusters of locations.
Results and Future Work
International visitors stay significantly longer (averaging ~49 days vs. ~4 days for locals) and visit more locations (~2.9 vs. ~1.3). Across almost all groups, the relationship $S^{rand} > S^{unc} > S^{weight}$ held true, meaning that when accounting for specific probabilities and distances, human mobility appears much more "ordered" than a simple count of locations would suggest.
While this study focuses on irregular movement in isolation, future work aims to develop a unified model that incorporates both regular (routine) and irregular (context-driven) features. This would allow for more robust simulations of disease spread or infrastructure stress during peak travel seasons.
Summary updated