A team of computational social scientists out of the Indiana University-Bloomington have stumbled upon every amateur stock market enthusiasts dream – a crystal ball that predicts the market with almost 90% accuracy 3 days ahead of time. It’s called Twitter.
The researchers accidentally found this data when measuring the overall public mood on Twitter using an algorithm they developed combining some of Google’s search data with an established psychiatric evaluation parameter known as the Profile of Mood states.
Using 72 specific words people use to describe how their feeling (nice, angry, abused, panicked, etc) adding to those similar words from Google’s databases, and marrying that to any tweets that start with any version of “I feel,” the researchers have created a relatively accurate barometer of public mood.
Similar projects have taken this task on in bits and pieces (Such as “We Feel Fine” and “Twistori”), but none have used the level of statistical analysis these researches did.
Before they realized the correlation to the Stock Market trend, they had analyzed several other events – how Tweets compare to emotion on Election days and holidays, for example. It was out of sheer curiosity that one researcher compared them to the stock market and found an 86.7% correlation with the stock markets rise and fall with the emotion (Twitmotion?) ‘calm.’ The researchers acknowledged that this isn’t the most accurate algorithm for stock prediction possible, but it’s still a surprising and important sociological result. Via Wired