This week, a group of scientists at MIT published a ground-breaking study in which they use Twitter to predict what people are tweeting. In the past, researchers have successfully used Twitter to predict stock market fluctuations, box-office sales, riots, and even typhoons, but this is the first time anyone has pursued a project on this scale. They achieved an impressive prediction accuracy of approximately 74%.
According to Professor Johnson Billing, head author of the study, this work has numerous promising applications. “If somebody had access to Twitter data but did not feel like looking at Twitter, they could run this algorithm to obtain a rough idea of the latest tweets,” said Billing.
Professor Billing and his students have received a 4 billion dollar grant from the U.
S. government to continue improving their algorithm, with the lofty goal of achieving 82% accuracy by 2015.
