Computer scientists use Twitter to predict UK general election result

May 7, 2015

Computer scientists from the University of Warwick are using Twitter to predict the outcome of the UK general election and believe their forecasts could be more accurate than traditional opinion polls.

They are working in collaboration with partners in the Department of Journalism at City University London and the Information Technologies Institute (ITI-CERTH, Greece).

Together, the team is using an algorithm that harvests political tweets, aggregates various features about every party and then injects this information into conventional polling reports, producing a daily prediction of voting share.

With the outcome of the in terms of seats more uncertain than ever, this approach will provide crucial insights into how public opinion is developing and what factors might be influencing any changes in support. Early results of the system have already tracked the surge in support for the SNP and the fluctuating fortunes of UKIP.

Details of the data produced so far, as well as commentary on the results and overview of the study, can be found here: www.electionprediction.eu.

Warwick researcher Adam Tsakalidis said: "We are trying to define and extract meaningful features out of the noisy, user-generated content published in Twitter. This includes the number of users mentioning a political party and the number of users who have expressed a negative opinion about this party on a certain day. We then put all this information into our forecasting model, along with the parties' share of the vote as measured by . Predicting elections using data has been tried in the past, with varying results. We have evidence from our previous work that our approach is very effective."

Adam, who has developed the core of the prediction algorithm, improved the model and tested it again in the Greek election in January, achieving better results than all 31 most recent polls leading to the election, as well as the three exit polls that were announced once the ballots closed.

He added: "We have now adapted our model for the UK general election and we are in a position where we can make a daily prediction and identify trends in for each party. One of the key features we are incorporating in the model is the sentiment conveyed in the tweets. We are particularly interested in automatically identifying the sentiment expressed towards specific politicians or parties and topics such as immigration. This will help us obtain more accurate predictions as well as better understanding of the reasons behind public support or discontent."

City University's Professor Steve Schifferes added: "Social media is now central to how people consume election news. This ground-breaking research will give us deeper insights into how people are developing their own understanding of the election that goes beyond the traditional news media's agenda, and could enable us to spot early trends which could affect the overall outcome."

ITI-CERTH researcher Symeon Papadopoulos noted: "Although social media has already been used for some time as a means of discovering current trends and as a proxy of the interests and perceptions of the public, research has only recently started exploring its potential for predicting real-world outcomes such as the results of elections. To date, there is no reliable approach for achieving this. Yet, our approach, which has already been tested in a couple of recent elections with great success, carries the potential of creating the basis for such a solution for the problem of election prediction."

Explore further: Election most tweeted event in US political history

More information: The model being used was successfully applied in three different countries in the EU 2014 elections, as part of theSocialSensor research project.

Related Stories

Election most tweeted event in US political history

November 7, 2012

Election day in the United States became the most tweeted about event in US political history Tuesday with enthusiastic netizens firing off 20 million poll-related tweets, the social network said.

Election surprises tend to erode trust in government

July 24, 2014

When asked who is going to win an election, people tend to predict their own candidate will come out on top. When that doesn't happen, according to a new study from the University of Georgia, these "surprised losers" often ...

Recommended for you

Inferring urban travel patterns from cellphone data

August 29, 2016

In making decisions about infrastructure development and resource allocation, city planners rely on models of how people move through their cities, on foot, in cars, and on public transportation. Those models are largely ...

How machine learning can help with voice disorders

August 29, 2016

There's no human instinct more basic than speech, and yet, for many people, talking can be taxing. 1 in 14 working-age Americans suffer from voice disorders that are often associated with abnormal vocal behaviors - some of ...

Apple issues update after cyber weapon captured

August 26, 2016

Apple iPhone owners on Friday were urged to install a quickly released security update after a sophisticated attack on an Emirati dissident exposed vulnerabilities targeted by cyber arms dealers.

1 comment

Adjust slider to filter visible comments by rank

Display comments: newest first

LagomorphZero
not rated yet May 07, 2015
Eventually these forecasts will be so accurate that we don't have to actually vote!

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.