Description
Twitter, the most popular micro-blogging site, having over 500 million registered users as of 2012 and creating over 340 million tweets per day, has caught the attention of socio-geographic researchers. With its availability from mobile phones, Twitter allows users to send messages with their geographic coordinates. Thus, with respect to the huge quantity and large diversity of crowds publishing tweets, massive valuable geosocial knowledge can be extracted, which can provide important implications for various applications, such as human geography, urban science, location-based services, targeted advertising, content delivery networks, and social media research. This study aims to uncover human behavior patterns, especially lifestyle of college students, by analyzing the spatio-temporal dynamics of geo-tagged tweets. Four Midwest college towns are selected as study area including West Lafayette, IN, Bloomington, IN, Ann Arbor, MI and Columbus, OH. First, overall distribution of tweets is revealed with point density tool, and individual frequent user’s spatial pattern is derived with Expectation-Maximization clustering. Then temporal analysis is performed with three time intervals: hour of a day, day of a week and month. Moreover, with land use data, further insight about where tweet incidents are can be discovered, leading to a deeper understanding about the population mobility and flow patterns of Twitter users. Finally, the space-time locations of tweet cluster, which are likely to be events, are identified with space-time statistics (STSS). This study suggests the potential of using geo-tagged micro-blogging service such as Twitter in understanding the human behavior patterns, and its possibility in socio-geographic research.
Start Date
11-2014
Document Type
Presentation
Keywords
GIS, social media, Twitter, human behaviour pattern
Session List
Lightning Talk
Included in
From Social Media Data to Human Behavior Patterns: Analyzing Spatio-Temporal Dynamics of Tweets in Midwest College Towns
Twitter, the most popular micro-blogging site, having over 500 million registered users as of 2012 and creating over 340 million tweets per day, has caught the attention of socio-geographic researchers. With its availability from mobile phones, Twitter allows users to send messages with their geographic coordinates. Thus, with respect to the huge quantity and large diversity of crowds publishing tweets, massive valuable geosocial knowledge can be extracted, which can provide important implications for various applications, such as human geography, urban science, location-based services, targeted advertising, content delivery networks, and social media research. This study aims to uncover human behavior patterns, especially lifestyle of college students, by analyzing the spatio-temporal dynamics of geo-tagged tweets. Four Midwest college towns are selected as study area including West Lafayette, IN, Bloomington, IN, Ann Arbor, MI and Columbus, OH. First, overall distribution of tweets is revealed with point density tool, and individual frequent user’s spatial pattern is derived with Expectation-Maximization clustering. Then temporal analysis is performed with three time intervals: hour of a day, day of a week and month. Moreover, with land use data, further insight about where tweet incidents are can be discovered, leading to a deeper understanding about the population mobility and flow patterns of Twitter users. Finally, the space-time locations of tweet cluster, which are likely to be events, are identified with space-time statistics (STSS). This study suggests the potential of using geo-tagged micro-blogging service such as Twitter in understanding the human behavior patterns, and its possibility in socio-geographic research.