Detecting Chinese stock information based on hidden Markov model.

The unobservable state of stock information was modeled based on Hidden Markov Model, and a transition probabilities matrix of information state was built to describe the dynamic association properties in temporal dimension. Based on 5-minutes high frequency trading data and using Bayesian inference and MCMC sampling, the information state and strength of Shanghai Stock Index and sample stocks of SSE 50 in August 2010 were estimated. Empirical results prove the model has effective ability to identify information, and show that the information effects have characteristics of aggregation in Chinese stock market. On the base of the estimated transition probabilities matrix of information state, it is surmised that the probability that an item of information was absorbed by the market after 100 minutes is 99% in Chinese stock market. [ABSTRACT FROM AUTHOR]