1 using gis to understand behavior patterns of twitter users yue li m.s. civil/geomatics engineering...
TRANSCRIPT
![Page 1: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/1.jpg)
1
Using GIS to Understand Behavior Patterns of
Twitter Users
Yue LiM.S. Civil/Geomatics Engineering
Purdue UniversityCommittee: Dr.Jie Shan (Chair), Dr.Nicole Kong,
Dr.James Bethel
![Page 2: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/2.jpg)
2
Introduction
• Volunteered Geographic Information (VGI)1
− Emergency management, event detection, tourist behavior, knowledge discovery…
− The most popular micro-blogging site
− Tweets with longitude and latitude
− A gold mine for scholars in geography, linguistics, sociology, economics, health, and psychology2
− Marketing, advertising, regulation,…
![Page 3: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/3.jpg)
3
Research Goal
• To discover the spatio-temporal pattern of tweets
• To infer the human mobility patterns behind the tweets
• To understand the lifestyle of college students
![Page 4: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/4.jpg)
4
Study Area
• College town/city, Big Ten Universities
• West Lafayette, IN
− Most densely populated city in IN
− Home of Purdue University
• Ann Arbor, MI
− University of Michigan
• Bloomington, IN
− Indiana University, Bloomington
• Columbus, OH
− Ohio State University
![Page 5: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/5.jpg)
5
Data
• Geo-tagged tweets downloaded with Twitter Streaming API
• With longitude and latitude at time of posting
• Nov 18, 2013 to April 2, 2014
− West Lafayette : 59,238
− Ann Arbor: 220,117
− Bloomington :247,202
− Columbus: 1,936,238
![Page 6: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/6.jpg)
6
Methods
• Pure Spatial
− Point density analysis
• Pure Temporal
• Spatio-Temporal
− Tweets in Land Use
− Event/Anomaly detection
− Individual twitter user patterns
![Page 7: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/7.jpg)
7
Tweets in West Lafayette
![Page 8: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/8.jpg)
8
Tweets in Ann Arbor
![Page 9: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/9.jpg)
9
Tweets in Bloomington
![Page 10: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/10.jpg)
10
Tweets in Columbus
![Page 11: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/11.jpg)
11
Tweets by Hour
![Page 12: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/12.jpg)
12
Tweets by Hour
![Page 13: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/13.jpg)
13
Tweets and Land Use
• Land use in Ann Arbor, MI− Industrial
− Mixed Use
− Office
− Public/Education
− Recreation
− Residential
− Transportation
− Vacant
• Spatially join the tweets with land use
![Page 14: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/14.jpg)
14
Tweets and Land Use
![Page 15: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/15.jpg)
15
1 - Commercial; 2- Industrial; 3- Mixed Use; 4- Office; 5- Public/Education;6 – Recreation; 7- Residence; 8- Transportation; 9- Vacant/River
Tweets and Land Use
![Page 16: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/16.jpg)
16
Event Detection
• Spatially and temporally aggregated
− Football game, concert, festival,…
• Use Purdue shooting on Jan 21, 2014 as an example
− Lockdown from around 12-14pm
• Temporally
− 710 tweets in 12-14pm Jan 21, 231 unique users
− 7443 tweets in 12-14pm in the whole datasets, 1080 unique users
• Spatially
− How to measure spatial anomaly?
![Page 17: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/17.jpg)
17
Hypotheses
• Challenge: Inhomogeneous/clustered process even outside lockdown period
− Were tweets more significantly clustered during lockdown than average?
• Intensity of tweets is correlated with distance to campus buildings
• Extent of clustering is positively correlated with chi-sqare value
![Page 18: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/18.jpg)
18
Covariate: Purdue Buildings
Purdue Building Shapefile converted to tesselation
R libraries: maptools, sp, spatstat
Functions: as.mask → im → tess
![Page 19: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/19.jpg)
19
Randomization Test
Algorithm (by Ken Kellner):
1. Select 710 random tweets from dates 1/16/14 - 1/26/14 and hours 12am - 14pm without replacement
2. Call quadratcount() and quadrat.test() on new random dataset
3. Save chi-square value
4. Repeat 1000 times to obtain distribution of chi-square values
5. Compare actual chi-square value obtained on 1/21/14 with distribution
6. Quasi-p value: proportion of values more extreme than obtained value
Assumption: greater chi-square value = more inhomogenous/clustered
Tested with simulation
![Page 20: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/20.jpg)
20
Randomization Test Result
Chi-square: 85162.85Quasi-p value: 0.038
• We were able to detect a change in the pattern of tweets during the lockdown, when presumably more people were stuck in Purdue buildings than average.
![Page 21: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/21.jpg)
21
Event Detection
• We can see anomaly from Twitter data both temporally and spatially
• However, we are still looking for a complete and integrated algorithm, and apply it to other events
• To be cont’d
![Page 22: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/22.jpg)
22
Frequent Twitter Users
• Top 10 Twitter users with the most tweets in Ann Arbor
• Plot the tweets of individual Twitter user
• Four typical patterns− Work-Home
− Work-Road-Home
− Work-Home-Short Visit
− Multiple Clusters
![Page 23: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/23.jpg)
23
Frequent Twitter Users
![Page 24: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/24.jpg)
24
Frequent Twitter Users
![Page 25: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/25.jpg)
25
Future Work
• On-going research
• Complete analysis in all 4 study areas, and compare the patterns
• Develop/Find an algorithm for event detection
• …
• Any suggestions are welcomed!
![Page 26: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/26.jpg)
26
References
• 1. Goodchild, M. F., 2007. Citizens as sensors: The world of volunteered geography, GeoJournal, 69, 211-221.
• 2. Ghosh, D., and R. Guha, 2013. What are we ‘tweeting’ about obesity? Mapping tweets with topic modeling and Geographic Information System, Cartography and Geographic Information Science, 40(2), 90-102.
![Page 27: 1 Using GIS to Understand Behavior Patterns of Twitter Users Yue Li M.S. Civil/Geomatics Engineering Purdue University Committee: Dr.Jie Shan (Chair),](https://reader036.vdocuments.us/reader036/viewer/2022062712/56649c895503460f94942d0b/html5/thumbnails/27.jpg)
27
QUESTIONS?