|iStockphoto / James Steidl|
Using a self-organizing fuzzy neural network model, researchers were able to correlate stock market movement 3 days in advance with a nearly 90% success rate by analyzing mood from a statistical sampling of tweets from Twitter.
Two mood measurement tools were used in the model. OpinionFinder measured public sentiment with simple positive / negative values while a new tool created by the authors called Google-Profile of Mood States, measured mood along six dimensions.
The paper discounts the Efficient Market Hypothesis (EMH) which states that, on average, returns greater than the market average can’t be obtained because prices reflect all information that is currently publicly available. However, the authors don’t take into account that while there may be a short-term opportunity to take advantage of such tools, once publicly available, these tools will themselves simply provide new sources of information that will be built into to market prices, thus reducing the window of opportunity for using them to an advantage.
A number of ways to increase the accuracy of the authors’ model can be imagined. For example, an improved model might give proportional weighting to the number of subscribers that an information source reaches to take in to account influence.