Intro
- location-based social media (LBSM) data are widely used for modeling human activity patterns
- research questions: 1) how do different activity space indicators change with different data collection durations? 2) are there substantial differences in terms of the activity spaces in these three cities?
Previous Work
Modeling human activity spaces: indicators and measurements
[external morphology of human activity space]
- minimum convex hull: sensitive to outliers
- alpha shape: less sensitive to outliers and can represent activity spaces more accurately
- ellipse-based measures: activity spaces are based on distances and directions from the activity center of an individual's visited locations, less sensitive to outliers than the convex hull
- radius of gyration (ROG): represents the spatial dispersion and activity range of individual daily activities, less sensitive to outlier points
[internal structure of activity spaces]
- density based: illustrates the visitation frequencies of activity location
- network-based: investigates the connections and links between these locations
- probability-based: explore the likelihood of visiting a location (e.g. entropy)
[Using location data to model user activity spaces]
- LBSM data sets cover a larger sample size and can easily be accessed by application programming interfaces
- LBSM data quality issues such as low resolution, completeness, and consistency will affect the analysis
Data and the study area
- they used the unique identifier (e.g. user account ID), the coordinates of check-in locations, and the timestamp of check-ins
- study areas are Beijing, Shanghai, and Guangzhou in China
Method
Define external activity space indicators
- minimum convex hull: defined as a polygon that contains all points and has no internal angles greater than 180 degrees on a two-dimensional plane, easy to compute but it is less accurate if the person's activity space is irregularly shaped
- alpha shape: generalization of the minimum convex hull, less affected by outliers
- standard deviational ellipse (SDE)
- radius of gyration (ROG)
Define internal activity space indicators
- entropy: represents the randomness of activity patterns, a higher entropy value represents a higher degree of randomness
- spaces minimum spanning trees (MST): subset of the edges of a connected, graph that connects all locations together without any cycles
- estimated home location (based on kernel density): transform a set of points into a continuous representation of density in space
Compare indicators and fit models
- 1) compared the results of the four external indicators in modeling the sizes of LBSM activity spaces
- 2) conducted a Wilcoxon rank sum test on the average size of activity spaces in the three cities
- 3) tested how different data collection durations affect the magnitude of the seven activity space indicators
Result
Comparison of activity space indicators
The impact of data collection duration on activity space indicators
- when using twelve-moth average data, the average entropy and the average MST distance are very close to the approximated limit value in all three study areas
