Study Of Stock Market Opportunism Based On The Hidden Markov Model

With the development of China’s economy,the per capita disposable income of residents continues to grow,and more and more investors invest funds in the stock market to achieve asset appreciation.However,the stock market has the characteristics of high volatility and high returns,so in order to achieve value investment,the general investment philosophy is "buy low and sell high",so it is particularly important to choose the timing of buying and selling stocks.Accurate predictions of stock prices can help investors find the right time to enter and exit the stock market.Historically,many scholars have studied the method of stock price forecasting by constructing models,which can be roughly divided into classical time series forecasting models and machine learning forecasting models,and as a kind of machine learning,the hidden Markov model(HMM)has been widely used by scholars in language recognition and processing,biology and other fields,but some scholars have introduced it into the financial field for stock price forecasting.However,the traditional hidden Markov model needs to be improved when predicting stock prices,so this paper introduces the traditional hidden Markov model,forecasting methods and forecasting processes,and analyzes the shortcomings of the model,so as to improve and improve it to achieve more accurate prediction of stock market price changes.At the same time,the forecast results are simulated trading,long and short strategies are constructed to backtest the predicted price,and the simulated investment results are evaluated.The main contents of this study can be divided into the following aspects:(1)Improve the index,process the original stock price data,and form a series of relative price fluctuations as the observation sequence of the model;(2)Demonstrate that compared with the stock price series,the series of ups and downs has higher stability and is more in line with the characteristics of Gaussian distribution;(3)Determine the optimal number of hidden states according to AIC and BIC criteria,and train the model to obtain model parameters;(4)The value range of the rise and fall is divided by equal width,and the method of integration is used to obtain the largest probability of rise and fall,and finally the predicted stock price is obtained according to the closing price of the stock on the day and the predicted series of rise and fall;(5)The accuracy and error of the model were evaluated by means of mean absolute percentage error(MAPE)and root mean square error(RMSE),and the error comparison and analysis of the traditional hidden Markov model and LSTM model were carried out.(6)Build a long strategy and a long and short strategy,backtest the prediction results of the model,and compare the simulated investment results.The innovation of this article is:(1)The traditional stock price series is relativized as the observation variable of the model,so that there is no need to use the historical similar likelihood method for matching when predicting stock prices,and the rise and fall series are more stable and conform to the Gaussian distribution than the stock price series;(2)The weighted value of the observation sequence is taken by the value of the state transition probability matrix,and the value range of the rise and fall series is divided equally,the probability of occurrence of each small interval is calculated by the method of integration,the largest probability value is used as the prediction result of the model,and the predicted price of the stock is calculated by combining the closing price of the stock on the day.