DATA MINING TWITTER TO PREDICT STOCK MARKET MOVEMENTS

In this paper we apply sentiment analysis of Twitter data from July through December, 2013 to find correlation between users’ sentiments and NASDAQ closing price and trading volume. Our analysis is based on the Affective Norms for English Words (ANEW). We propose a novel way of determining weighted mood level based on PageRank algorithm. We find that sentiment data is Granger-causal to financial market performance with high degree of significance. “Happy” and “sad” sentiment variables’ lags are strongly correlated with closing price and “excited” and “calm” lags are strongly correlated with trading volume.