Costa Rica News – Move over, political pollsters. Arnoldo Müller-Molina has a new way of using Twitter to predict election results.
Müller-Molina, a native of Costa Rica, moved his data analysis company simMachines to St. Louis in 2012 after receiving a $50,000 Arch Grant. Five days before his home country’s presidential election on Feb. 2, he fed 12,455 tweets about Costa Rica’s various political parties into simMachines’ software.
The software looked for keywords to categorize tweets as positive, negative or neutral, and then it ranked the parties. The Citizens Action Party led by Luis Guillermo Solis had been running fourth in traditional opinion polls, but it showed up as No. 1 in Müller-Molina’s analysis.
When the votes were tallied on Feb. 2, Solis had a narrow lead over presumed front-runner Johnny Araya. Reuters reports that the two will face each other in a runoff election in April.
Müller-Molina says companies can use the same methodology to analyze how social media users are reacting to their products. And he sees more applications in the political sphere.
“If we could identify these minor discrepancies in traditional polling there, then the same technology can be applied everywhere,” he said in a news release. “We showed how machine learning could promote democracy and protect the will of the people.”
By David Nicklaus, [email protected]
David Nicklaus is business columnist at the St. Louis Post-Dispatch