Stock Price Index Prediction Based On Hidden Markov Model

Stock price is uncertain.If investors can predict the price trend beforehand,they will be able to avoid the risk of price fluctuation,reduce investment losses and even get excess returns.With the upsurge of large data,more and more statistical learning models are applied to the stock price prediction,and the Hidden Markov Model is one of them.The Hidden Markov Model is developed on the basis of the Markov chain,and it is used to study a set of hidden states.The model is a double stochastic process.It consists of two parts:Markov chain and general stochastic process,which are used to describe the relationship among states and the relationship between states and observed values.In this paper,we predict the stock price and the stock market states based on the Hidden Markov Model.We take the stock price index-CSI300 index as the research object,and we discuss the feasibility of the related theories of the Hidden Markov Model in the price prediction and the stock market prediction,then we build a stock price prediction model that is suitable for the national conditions of our country.The empirical research mainly includes the steps of data selection and inspection,determining the number of hidden states,parameter estimation and prediction.And we improve the continuous Hidden Markov Model from the three aspects-optimizing the model input based on artificial neural network algorithm,optimizing the initial value of the model based on ISODATA algorithm and predicting the stock price based on multi-day weighted prediction method,thereby we put forward an improved Hidden Markov Model.The improved Hidden Markov Model overcome the shortcomings of the basic model which can not predict the specific price data,and it can predict the market status and stock index price.Compared with the basic model,the prediction error of the improved model is greatly reduced,and the MAPE value decreases from 1.24 to 0.75.The empirical results show that this model can predict the state of the stock market and the stock price index,the stock market in a state(bull market or bear market)will last for a period of time,and turning from one state to another is difficult.Besides,the improved model can increase the prediction accuracy,which can make significance to the actual investment.