Researchers have developed a new statistical model that simulates human mobility patterns, mimicking the way people move over the course of a day, a month or longer.
The model, developed by scientists at North Carolina State University and the Korea Advanced Institute of Science and Technology (KAIST), is the first to represent the regular movement patterns of humans using statistical data. The model has a host of potential uses, ranging from land use planning to public health studies of epidemic disease.
The researchers gave global positioning system (GPS) devices to approximately 100 volunteers at five locations in the U.S. and South Korea and tracked the participants' movements over time, according to study co-author Dr. Injong Rhee, a professor of computer science at NC State. By plotting the points where the study participants stopped, and their movement trajectories, researchers were able to determine patterns of mobility behavior.
For example, Rhee says, the researchers found that people tend to perform multiple activities in clusters that are in close proximity to each other – such as going to a bank, a dry-cleaner and a pharmacy that are all located on the same street. Furthermore, the researchers found that the study participants tend to more frequently visit locations that are popular among other people.
These behaviors illustrated statistical patterns. For example, Rhee explains, people will try to make the most efficient use of their time and effort by clustering activities together that are in geographical proximity to each other. This behavior creates patterns in which people make many short "jumps" within the clustered areas while making a few long jumps among the clustered areas. These patterns are best explained by statistical processes called self-similar points of visits and power-law distribution of jumping distances.