News – Study Finds Twitter May Be Much More Than Random Talking

Twitter seems an unlikely place to garner anything more than short updates, a handful of favorite current links from friends, and a way to share your life with the world in a 140 characters or less.  However, a group of analysts have taken the time to study what may be a very important gauge of the stock market buried within the millions of tweets throughout the site.

That’s right, I said the stock market.  For years, people have been trying to formulate ways to predict the actions of the market with any degree of consistency, only to find that their predictions are about as accurate as the flip of a coin.  But now, by using a scale of emotions in the form of an algorithm and a pool of nearly 10 million tweets from 2008 to present day, they can now predict the swing of the Dow Jones about 87 percent of the time.

The algorithm is called GPOMS, or Google-Profile of Mood States, and it focuses on six emotional states: happiness, kindness, alertness, sureness, vitality, and calmness.  The final state, calmness, is the emotional state that has been shown to best predict the Dow Jones.  They measured the so-called “calmness index” against the Dow Jones rises and falls on a daily level and found that there is a pretty defined correlation between the two.  So much so, that they are continuing their study because they feel that they can routinely predict the attitude of the market 6 days in advance.

Going forward, they will need to separate international tweets from American ones in order to gain a more accurate reading, as well as take into account the fact that the correlation, while defined, may be broken into many sub-categories that combine to create their result.

Either way, it’s interesting to see that something as cut and dried as Twitter seems to be may actually be the perfect tool to measure human actions and thoughts, and provide researchers with a sort of Rosetta Stone for deciphering the attitudes of the masses.

To read the original article, please go to http://www.technologyreview.com/blog/arxiv/25900/?p1=Blogs

Comments are closed.