Toxic fumes

Toxic fumes: With pollution being a major concern in London, analysts were interested in Yeonsoo's findings

Twitter could become an early warning system for outbreaks of pollution, according to a Warwick Business School graduate who took her idea to Westminster City Council in London.  

With the health impacts of toxic fumes being a major concern in the capital, Yeonsoo Jang was invited last November to present her dissertation findings to a group of data analysts involved with the ‘clean tech city’ arm of Westminster City Council’s Smart City Programme.  

In her dissertation, Yeonsoo set out to investigate whether public reactions to London air quality posted on Twitter could be used as an additional source of air quality monitoring. 

Collecting 246,842 tweets posted between July 2018 and June 2022, Yeonsoo then applied techniques learned during her MSc Business Analytics course such as using artificial intelligence tool VADER (Valence Aware Dictionary for Sentiment Reasoning) to analyse the sentiments expressed in them about air quality.  

The keywords and phrases she was looking for were arranged around themes such as ‘London air quality’ and locations like Soho, Westminster and the Edgware Road. Although London has a world-class air quality monitoring network, its fixed monitoring stations cannot cover the whole urban space. Yeonsoo’s supposition was that Twitter posts might be an indicator of pollution levels in the areas without monitoring capabilities.   

“During the 2016 Presidential election in the US, many data scientists and data geeks built their own Twitter sentiment analysis models,” she said. “I believe that sentiment analysis is one of the simplest and most efficient ways for governments and the general public to interact in real-time.”  


So it proved. After comparing the correlations between her Twitter data and air quality indexes, she found that there were, indeed, strong societal reactions to different pollutants such as nitrogen dioxide, sulphur dioxide and particulate matter over time.  

Moreover, social media reflected air quality conditions better than radio, television and newspapers, raising the question whether mainstream media should take London air quality more seriously as an issue.  

“The Westminster analysts were really interested not only in my findings but also in my methodologies such as the data techniques and models I used,” Yeonsoo said.  

Given that London’s sparsely located monitoring stations are expensive to maintain, and that forecasting models are used to provide air quality information in areas without monitoring stations, Yeonsoo believes that an additional data source that is affordable could be the way forward.  

Providing precise and real-time data, Twitter could also be used in cities with inferior air quality monitoring systems.  

“Social media as a data source is a great way of scaling data at minimum cost,” the Warwick alumna said.  

Pietro Micheli, Professor of Business Performance and Innovation at Warwick Business School, who supervised Yeonsoo during her dissertation, said: “This is a demonstration of how our MSc course in Business Analytics can effectively equip students with the ability to use and analyse large data sets and apply their skills to current business and societal problems.”