Prediction Of Stock Price Based On Hidden Markov Model

Focusing on the uncertainly of stock price, a random model of prediction on stock price on stock market was given by using hidden Markov Chain, which is a doubly stochas-tic process. By transition matrix or Gaussian mixture distribution, the two processes can describe the hidden status and observations. The model can connect the future status of stock price with historical status. Then we have to estimate the parameters of the model by historical datum. We now know the distribution of hidden status and observations. So we can predict the hidden status.In this paper, we apply the hidden Markov model to predict the future status of stock price. The hidden status is the future status of stock price. When we use the continuous model the observations are the index of stock price. When we use the discrete model the observations are the discretization of them. By the historical market we can estimate the parameters of the model, with the parameters and the present observations, we can predict the future status of stock price. To verify the effectivity of the model, we compare the prediction and the reality, and finally get the conclusion that given the same data continuous hidden Markov model is more effective than discrete hiddern Markov model with greater than 50% accuracy, ie 52.7%. Using discrete hiddern Markov model the accuracy can not even reach 50%. In general, although continuous hidden Markov model is more complicate and the requirement about data is more harsh, continuous hidden Markov model is more effective than discrete hiddern Markov model.