Abstract:

We collect Web financial text information from financial BBS（Bulletin Board System) and classify all these text information according to their word count. Then we do empirical analysis by factor model and GARCH(Generalized Autoreg Ressive Conditional Heteroskedasticity) models. Our three findings are: number of new posts and stock yield are directly related; posts with more words are expected to have greater affects on stock yield; the correlation between count of new posts and stock volatility is statistically significant.

Key words: Web text mining, stock market, factor analysis