Statistical Financial Physics Modelling And Financial Market Stochastic Forecasting System

Financial market is an open and complex dynamic system that can be considered a dynamic system of violent fluctuations which made up of a large number of interacting elements.The research on the fluctuation behavior of the financial market,the related statistical analysis and the prediction of the financial time series have always been the focus of.the economic and financial research fields.In particular,with the development of "Econophysics",more and more references have put forward micro price model to reveal the fluctuation of the stock price.In this paper,we employ the intrinsic mecha-nism of statistical physical system(Potts dynamic system)to characterize information interaction among investors in financial market,so as to establish corresponding price models.On this basis,the statistical characteristics of the simulated data generated by the model are discussed,and the rationality and effectiveness of the price model are verified by comparing with the real market data.In addition,another research content of this paper is based on artificial neural network,combined with stochastic theory,principal component analysis,empirical mode decomposition and so on.Several new prediction models are constructed,and then the prediction and error analysis of differ-ent trading data in different financial markets are made.The main structure of the this thesis is as follows:The first chapter introduces the research background and the current research status at home and abroad.Chapter 2 mainly introduces the modeling of the financial price series.On the ba-sis of the dynamic theory of the Potts model,Potts financial price model is constructed by using the modeling theory of financial market trading market operation mechanism.Potts dynamic system is a very important statistical physical system.The interaction of dynamic interaction particles in the system is used to describe the change of informa-tion transfer and investment attitude among different investment groups in the financial market.In chapter 3 mainly introduces the different statistical characteristics of the return time series and return interval sequence simulated by the Potts financial model.And defines new complex statistical methods such as MCID,q-MCID and EMD-MCID to study the characteristics of multiscale correlation between different return time series.Then the chaos method is introduced to compare the return time scries of the Potts fi-nancial model with the real market time series.Finally,the complexity and multifractal correlation of the return interval time series simulated by Potts model are studied and explored.Chapter 4 mainly introduces a new predictive model PCA-STNN,which is the combination of principal component analysis method(PCA)and stochastic time ef-fective neural network.Among them,the use of drift function and random Brownian motion to describe the stock market time intensity behavior,in order to maintain the original tendency of the principle of random movement in the forecasting process.In chapter 5 mainly introduces a new prediction model,that is hybrid EMD-STNN,which adds the empirical mode decomposition method(EMD)to the stochastic time effective neural network model that we introduced earlier.The multi-scale complex-ity distance analysis method is applied to the error analysis,and the accuracy of the prediction results is verified by testing the similarity rules with the original data.In chapter 6 mainly introduces a stochastic recurrent neural network,ST-ERNN,which combining the theory of multilayer perceptrons and Elman recurrent neural net-work,adding the stochastic time effective function to build the new forecasting model.In order to demonstrate the prediction of the new model ST-ERNN,this chapter con-ducts an empirical study on the price series of the crude oil and the stock price sequence respectively.Chapter 7 summarizes the innovation and main conclusions of this article.