Sunday, January 16, 2011

Twitter Predicts The Stock Market?

There is no longer any need to wonder if the DOW is heading up or down. Investors no longer need to hedge their investment risk with a diversified portfolio, now they need only follow Twitter.

According to Johan Bollen (Indiana University), the direction of the Dow Jones Industrial Average (DJI) can be predicted by following the general "mood" of Tweets on twitter.

He's not talking about tweets of professional stock pickers, or even amateur stock pickers for that matter. In fact, he's talking about just the general mass of tweets. Apparently the method consists of scanning tweets for certain key words or phrases that denote emotional states of happiness or sadness or fear, etc..

Based on the overall consensus of these anonymous tweets, he's able to predict the direction of the Dow within 3 days away. What's more, he claims he's got an 87% accuracy rating with the method!

Here's an interview with Mr. Bollen about a paper he wrote detailing the method.

He has an interesting premise, but there are serious problems with the data. For example, as the interviewer points out, the population of Twitter is demographically skewed to the younger set. Many of that generation are not interested in investing or finance in general. That doesn't matter, according to Bollen. He says that his model is accurate even when using mundane "I did lousy on that calc exam" kind of tweets!

Beside the demographic skew, Twitter has only been in existence since 2006 and through much of that time it was relatively small. It wasn't insanely popular until just a few years ago so the sample set of data and the history of the model's track record is very small - just over 2 years!

87% accuracy in the stock market over the last 2 years is not enough to really prove success, in my opinion.


Post a Comment