Research On Stock Market Analysis Methods Based On Evidential Reasoning And Hierarchical Belief Rule Base

The stock market is an important place for investment transactions and is known as the " thermometer" of the national economy.However,the stock market is a complex dynamic system with high risks and high returns,and is characterised by non-linearity,randomness and instability.It is difficult for amateur investors who lack professional investment knowledge to make clear and accurate judgements.At the same time,investors often refer to multiple types of technical indicator data(such as the opening price,the highest price,etc.)to dig into the pattern of price fluctuations to achieve investment decisions,but the ability to manually process these complex data is limited.Therefore,intelligent analysis of stocks is necessary to provide an effective reference basis for non-specialists.Although the intelligent stock analysis method has made some achievements,there are still some problems to be solved.On the one hand,most intelligent methods analyse stocks from the perspective of quantitative index data,and ignore the qualitative information existing in the analysis process.On the other hand,most of the intelligent model outputs rely on massive data training,which are difficult to be explained and verified by relevant financial theories,and suffer from low accuracy and poor interpretation.To solve the above problems,this paper proposes a model based on evidential reasoning(ER)and hierarchical belief rule base(BRB)to analyse the stock market,and divides the research process into three stages: trend prediction,market assessment and investment decision,with the following main contributions.(1)A stock price prediction model based on principal component regression(PCR)and HBRB is proposed to address the problem of reduced model accuracy and efficiency due to multiple indicator inputs.PCR is used for indicator screening to solve the problems of overfitting and accuracy degradation caused by high-dimensional data input.Secondly,the HBRB-based trend forecasting model is constructed based on the screened indicators.The forecasting process of this model is relatively transparent and substantially increases the belief level of the forecasting results.Finally,the model parameters are optimised by projection covariance matrix adaptation evolutionary strategies(P-CMA-ES)to further improve the accuracy and reliability of the model.(2)A stock market analysis model based on Evidential reasoning(ER)and a hierarchical belief rule base is proposed to address the problems of inability to integrate qualitative and quantitative information and the lack of financial theory support for the model.The model is divided into two aspects.Firstly,it is an ER-based market assessment model,which is able to combine financial mechanisms with ER algorithms and integrate different indicator data to deal with different types of information,both quantitative and qualitative,in an integrated manner.The second is the HBRB-based integrated decision model,which draws on a large amount of expert knowledge and experience in the model construction process to ensure the accuracy and reliability of the model.In addition,the model allows investors to clearly understand the reasoning process of each step,helping them to make scientific and reasonable investment decisions.(3)Based on the stock analysis model of ER and HBRB,a stock market analysis prototype system was designed and developed to provide an application reference for utilising the algorithms in this paper.In summary,this paper firstly takes a single stock as the research object and constructs a trend prediction model based on PCR-HBRB to predict individual stocks to help investors seize the best time to buy and sell,which can then ensure that the original information is not destroyed on the premise of reducing the dimension of stock data and improving the model calculation efficiency.Secondly,the SSE index is used as the research object to build an evaluation decision model based on ER-HBRB.The analysis of the index is conducive to determining the overall market situation and grasping the investment direction.This model incorporates investment theory and expert knowledge to give theoretical meaning to the output decision results and enhance model interpretability.Finally,this paper combines the algorithm to develop a stock analysis system to implement the theory into practical applications.