Stock Selection Model Based On HMM Mode And Its Application

The price of securities is affected by extremely complicated factors.Generally,when analyzing the trend of securities prices,it is necessary to analyze the market trend,industry trends and individual stocks in multiple stages and polymorphisms.Therefore,the conditional random field theory can be considered to be suitable for solving such financial problems.As a special conditional random field model,HMM has obvious advantages over other pattern recognition tools in describing the dynamic changes of the process,so that this paper attempts to explore the use of the HMM model to solve the stock selection problem in stock investment.Firstly,the index stocks with a positive weekly yield are named“up”stocks in this paper,with the others named“down”stocks.Then we select?closing price-opening price?/?opening price?,?highest price-opening price?/?opening price?,?opening price-lowest price?/?opening price?,daily turnover rate,?closing price-yesterday closing price?/?yesterday's closing price?and circulation market value as six observations.In the following step,using observation time series of the first ten days to train the HMM models for the“up”and“down”stock time series respectively,we gain the“up”model HMM₁ and“down”model HMM₂.Then we make use of the trained models HMM₁ and HMM₂ to predict the stock new observation time series likelihood value,which are the stock's“up”stock picking factor value y₁ and“down”stock picking factor value y₂ respectively,and comprehensively design the“up”conditional stock selection factor value y.In order to test the validity of the stock selection factor,after testing and comparing the information coefficients?IC?of multiple"stock selection factors",the paper selects the IC's significant"stock selection factor"as the final stock selection index.Then the top stocks with this“stock selection factor”value are selected as the investment portfolio,which is applied in market neutral and industry neutral quantitative hedging strategy design.In the following step,in order to further improve the investment portfolio,the paper introduces“entropy”to measure the consistency heat of“up”and“down”of the industry,constructing a quantitative hedging industry biased strategy based on industry entropy value allocation.The paper empirically analyzes the time series data from January 1,2016 to August 30,2018.Firstly,we collect the 55-day stock observation data of the Shanghai and Shenzhen 300Index constituent stocks?10 groups of 10 days six eigenvalue time series observation data and matching weekly rate of return data samples?to train the HMM model.Then the result of a statistical test on the IC of these"stock selection factors"displays the IC mean of the“up”conditional stock selection factor is significant.Then the value of this factor is selected as the investment portfolio in the top stocks in the industry as an industry neutral quantitative hedging investment strategy,which achieves a considerable revenue effect with the historical investment effect simulating by the rolling propulsion model.In order to improve the investment income,an industry-based neutral investment strategy based on industry entropy allocation has been established.After the historical test,it achieves better annualized rate of return and higher Sharpe on the basis of the original strategy,which fully demonstrates that the quantitative hedging strategy based on industry entropy value allocation by using the HMM model possesses exploration value on theoretical research and guiding value in financial investment practice.