Computer Science and Software Engineering Capstone Presentations
Fall Quarter
December 18, 2020
|
Derek Bui "Behavioral
Analysis of Citizen's Mobility during Pandemic: Instagram Case Study" (UWB CSS Faculty Research) Faculty Advisor: Dr. Afra Mashhadi |
|
Abstract This Capstone is a Data Science project focused on using
social media analysis for improving park and recreation usage to be fairer
and more inclusive. In particular this project will harvest information on
the magnitude and context of the park usage from social media platforms and
draw a comparison of recreational activities during and pre-pandemic. We seek
to understand how people sought after nature during time of lockdown and
social distancing and to what extent different segments of the population had
equal access to nature during this time. Our initial objective coming into the research
project was to find any correlation between green space visitation and a
pandemic. I used QGIS to work with the shapefiles for the parks of the 5
largest cities in the USA to find the 10 biggest parks in terms of area. I
developed a python script to crawl instagram data
in a JSON format as well as a time series analysis on R to plot the park
visitations over time. Our secondary objective was to extend this
functionality to other social media platforms as well as incorporate
sentiment analysis. For this, I developed an R app that would crawl and
analyze data from Twitter using their API. Once the scripts were developed and running, the
next step was to analyze the given data and look for any information or
trends we might see. Graphs and plots of different parks were pulled up and
divided into urban and peri-urban. Twitter sentiment analysis was documented
daily to show trends for keywords like "UW", "Election",
and "Covid". The approach found that in some of the larger
cities, there has been an increase in park visitation. Most notably Houston,
we can see that park visitation has nearly doubled from an average of 10-50
posts a day pre-covid to 20-100 a day during covid. The sentiment analysis on twitter showed that closer
to the election, certain words like "Covid"
and "Election" was mostly negative and
fear. As we got further and further away, "Election"
has mostly leaned towards negative and anger whilst "Covid" has shown positive trends. The results were in-line with
what was expected coming into this research project. |
|
|
Updated December 15, 2020