Predicting asset value through twitter buzz

Abstract

This paper describes early work trying to predict financial market movement such as gold price, crude oil price, currency exchange rates and stock market indicators by analyzing Twitter posts. We collected Twitter feeds for 5 months obtaining a large set of emotional retweets originating from within the US, from which six public opinion time series containing the keywords "dollar% t ", "$% t ", "gold% t ", "oil% t , "job% t " and "economy% t " were extracted. Our results show that these variables are correlated to and even predictive of the financial market movement. Except "$% t ", all other five public opinion time series are identified by a Granger-causal relationship with certain market movements. It is demonstrated that daily changes in the volume of economic topic retweeting seem to match the value shift occurring in the corresponding market next day. © 2012 Springer-Verlag Berlin Heidelberg.