Predicting stock market index trends through various attributes of Twitter data
MetadataShow full item record
We focus on correlating micro-blogging activity with stock-market indices like Dow Jones, NASDAQ, S&P500; and VIX. Specically, we collect microblogging messages related to stock market by using the appropriate hashtags, and we search for correlation patterns between stock-market indices and dierent aspects extracted from the above mentioned microblogging messages (also known as, tweets). These features measure the overall activity as well as the impact of a given message on the micro-blogging platform (Twitter), such as number of tweets, number of re-tweets, number of users, extent of reach of the tweets etc. We present detailed experimental results measuring the correlation of the stock market indices with the abovementioned aspects, using Twitter as our data source. Our results show that the strongest correlation exists between the number of tweets, extent of reach of any tweet as well as the impressions generated by that tweet, and stock market indices. The correlation gets stronger when the strong predictor variables are supplemented by not-so-strong predictor variables and analyzed using a multiple variable regression model. Thus we show that even relatively not-so-strong correlations between certain aspects of microblogging messages (or tweets) and stock market indices can be exploited to predict stock market indices that outperforms other baseline prediction strategies.