MODELING TWITTER SENTIMENT AS A FUNCTION OF PARTICULATE MATTER 2.5 FOR COMMUNITIES IMPACTED BY WILDFIRE ACROSS MONTANA AND IDAHO

Author

Matthew KellyFollow

Year of Award

2020

Document Type

Thesis

Degree Type

Master of Science (MS)

Degree Name

Computer Science

Department or School/College

Computer Science

Committee Chair

Douglas Brinkerhoff PhD

Commitee Members

Douglas Brinkerhoff PhD, Erin Landguth PhD, Jesse Johnson PhD

Keywords

twitter, pm2.5, sentiment, scraping, mining, autoregression

Subject Categories

Data Science

Abstract

Fine particulate matter (PM2.5) is a known pollutant with clinically detrimental physiological and behavioral effects. We consider Twitter sentiment as a potential indicator for well-being in communities impacted by wildfire-associated PM2.5 across Montana and Idaho spanning 5 years (2014-2018). From these geospatial air quality data and geo-tagged tweets, we trained county level models to examine the power of Twitter sentiment as a function of PM2.5. For all 24 counties sampled, we found between 1 and 8 affective dimensions where a positive �� 2 was detected with a significant F-statistic (�� < 0.05). Specifically, we show that sentiment for anticipation in the wildfire-prone county of Missoula, MT yielded respective training/test set �� 2 of 0.0958 and 0.0686 with a p-value for the F-statistic of 3.09E-07. These analyses support social media sentiment as a potential public health metric by showing one of the first observations of a relationship between PM2.5 and Twitter sentiment.

Recommended Citation

Kelly, Matthew, "MODELING TWITTER SENTIMENT AS A FUNCTION OF PARTICULATE MATTER 2.5 FOR COMMUNITIES IMPACTED BY WILDFIRE ACROSS MONTANA AND IDAHO" (2020). Graduate Student Theses, Dissertations, & Professional Papers. 11689.
https://scholarworks.umt.edu/etd/11689