Source-linked AI summary
Explore Spatiotemporal and Demographic Characteristics of Human Mobility via Twitter: A Case Study of Chicago
Feixiong Luo, Guofeng Cao, Kevin Mulligan, Xiang Li
TL;DR
The paper asks how demographic background affects human mobility, a question enabled by location-based social media’s combination of movement footprints and user information. It analyzes geo-tagged Twitter trajectories, derives homes and activity centers, infers demographics from names, and compares Chicago users across race/ethnicity, gender, and age. Mobility measures generally follow power-law patterns, while demographic differences—especially race/ethnicity—significantly affect urban mobility patterns.
Problem
The study addresses limited demographic consideration in human mobility research by examining how background factors relate to spatiotemporal mobility.
Method
The paper constructs Twitter user trajectories, detects homes and activity centers, and uses profile-name analysis with residential context to infer demographic groups for Chicago users.
Results
Human mobility measures generally follow power-law distributions, while race/ethnicity has the largest observed impact on mobility patterns among the studied demographic factors.
Takeaways & Limitations
Location-based social media support fine-scale mobility analysis that incorporates demographic information from user profiles and messages.
Takeaways & Limitations
The study focuses on Chicago and assumes users remain at a sampled location until a new activity is posted.
Abstract
from arXiv · showhide
Characterizing human mobility patterns is essential for understanding human behaviors and the interactions with socioeconomic and natural environment. With the continuing advancement of location and Web 2.0 technologies, location-based social media (LBSM) have been gaining widespread popularity in the past few years. With an access to locations of users, profiles and the contents of the social media posts, the LBSM data provided a novel modality of data source for human mobility study. By exploiting the explicit location footprints and mining the latent demographic information implied in the LBSM data, the purpose of this paper is to investigate the spatiotemporal characteristics of human mobility with a particular focus on the impact of demography. We first collect geo-tagged Twitter feeds posted in the conterminous United States area, and organize the collection of feeds using the concept of space-time trajectory corresponding to each Twitter user. Commonly human mobility measures, including detected home and activity centers, are derived for each user trajectory. We then select a subset of Twitter users that have detected home locations in the city of Chicago as a case study, and apply name analysis to the names provided in user profiles to learn the implicit demographic information of Twitter users, including race/ethnicity, gender and age. Finally we explore the spatiotemporal distribution and mobility characteristics of Chicago Twitter users, and investigate the demographic impact by comparing the differences across three demographic dimensions (race/ethnicity, gender and age). We found that, although the human mobility measures of different demographic groups generally follow the generic laws (e.g., power law distribution), the demographic information, particular the race/ethnicity group, significantly affects the urban human mobility patterns.
1 Introduction
The paper addresses the need to understand demographic influences on human mobility by combining location-based social media footprints with profile-derived background information. It proposes using Twitter name analysis to examine spatiotemporal mobility differences among demographic groups.
- Human mobility research supports traffic forecasting, disease-spreading studies, urban planning, and smart-city science.
- Traditional travel surveys are costly, time-consuming, and often limited to small samples, whereas Web and mobile technologies provide fine-grained data at larger scales.
- Existing mobility studies often emphasize spatial and temporal patterns while overlooking demographic and other background factors.
- Location-based social media provide user locations, profiles, and message contents, enabling individual-scale analysis of spatiotemporal mobility alongside demographic information.
- The paper applies name analysis to Twitter profiles to detect demographic groups and investigates their effects on users’ spatiotemporal mobility.
2 Methodology
The methodology constructs user space-time trajectories from geo-tagged Twitter data, filters sparse trajectories, derives mobility measures and homes, and infers demographics from names and residential context.
- The workflow comprises Twitter collection, trajectory construction, local-resident identification, demographic name analysis, and mobility investigation.
- Six months of geo-tagged tweets collected in the contiguous United States comprise over 300 million records from over 3 million users.
- Space-time trajectories: Each user’s posts are organized as a space-time trajectory, with the user assumed to remain at a sampled location until posting a new activity.
- Trajectory filtering: Trajectories with fewer than 24 tweets over six months are filtered out because their samples are too sparse to approximate movement profiles.
- Mobility measures: Radius of gyration measures activity range, while DBSCAN identifies frequently visited activity centers without prespecifying their number.
- Home detection: Home is defined as the most frequently visited activity center during 20:00–08:00, supporting resident identification and demographic analysis.
- Demographic inference: BISG combines surname information with census-tract demographics to infer race or ethnicity, while forename databases support gender and age estimation.
- Forename analysis: Forename age groups are assigned from birth-year patterns, with each group receiving an empirical probability based on occurrence fractions.
3 Analysis and results
Chicago Twitter users show regular spatiotemporal mobility patterns, while demographic differences—especially race/ethnicity—correspond to distinct residential distributions and mobility scales. Gender and age groups generally retain similar distributional forms, with narrower differences in selected mobility measures.
- Spatiotemporal characteristics: A correlation coefficient of 0.44 links detected Twitter residents across Chicago census tracts with the associated censused population.Twitter-user home density is higher in northern Chicago, especially near Lake Michigan, and closely represents the 2010 population distribution.
- Spatiotemporal characteristics: About 20,000 tweets were posted daily, averaging approximately 3 tweets per user per day, with peaks around 13:00–14:00 and 20:00–21:00.Tweeting was highest from roughly 10:00 to 22:00 and lowest around 4:00–5:00; weekends differed modestly from working days.
- Human mobility measures: The overall gyradius distribution follows a power law, with mean 183.03 kilometers and median 20.56 kilometers across three spatial scales.The three segments span local Chicago travel, regional travel from 25 to 1000 kilometers, and national travel from 1000 to 2000 kilometers; most users travel locally.
- Human mobility measures: Gyradius increased from 70.61 kilometers in January to 183.04 kilometers in June, whereas activity centers rose steadily from 1.4 to 5.4 without convergence.The activity-center distribution has mean 5.4 and median 4; over half of users have 2–4 centers and fewer than 10% have more than 10.
- Demographic differences: Race/ethnicity also corresponds to stronger residential separation and different activity-center spatial patterns, while gender and age groups show comparatively similar mobility distributions.African-American users tend to live in southern and western neighborhoods; White and Asian users are more spatially concentrated, and users over 60 show a somewhat higher share of 2–4 activity centers.
- Demographic differences: Race/ethnicity groups share roughly power-law gyradius distributions, but White and Asian users tend to have larger gyradii than African-American and Hispanic users.The reported mean gyradii are 275.56 km for White users, 85.13 km for African-American users, 84.89 km for Hispanic users, and 313.92 km for Asian users.
4 Discussion and conclusion
Using geo-tagged Twitter data, the study characterizes Chicago human mobility across space, time, and demographic groups. Mobility measures generally follow power-law patterns, while race/ethnicity, age, and gender show differing impacts, with race/ethnicity strongest.
- Study scope: Geo-tagged Twitter posts support analysis of Chicago users’ spatiotemporal and demographic mobility characteristics.The study selects users with detected homes in Chicago and analyzes their mobility using radius of gyration and activity centers.
- Overall mobility patterns: Human mobility measures generally follow power-law distributions across the analyzed Twitter users.This pattern is consistent with findings from cell-phone, transportation, and credit-card data.
- Spatial patterns: Activity-center distributions within Chicago align with socioeconomic development, while out-of-city distributions reflect Chicago’s socioeconomic links with surrounding areas.
- Demographic differences: Mobility measures generally retain power-law distributions across demographic groups but differ across race/ethnicity, gender, and age.The study derives demographic groups by analyzing names in Twitter user profiles.
- Demographic differences: Race/ethnicity has the largest apparent impact on mobility patterns, followed by age, while gender has the least.
- Implications and scope: The workflow and methodology can be generalized to other geographic areas and scales, while additional factors such as occupation and health status could be mined from social-media content.The authors identify bias in location-based social-media data as a recognized issue.