The investors’ neighbor herd behavior model in the financial market is a mathematical modeling process that combines psychology and financial behavior.The four circuit breakers in the U.S.stock market in 2020 and the "stock disaster" in the Chinese stock market in 2015 all reflect that the investors’ neighbor herd behavior of the stock market has exacerbated the abnormal market volatility.Investor’s neighbourhood herd behavior is formed under the complex internal and external factors of the stock market.Herd behavior have a variety of random and ambiguity features,neither conducive to the stability of the market,but also promote the formation of market bubbles.Due to the initiative,self-organization and heterogeneity features of the investor,it is necessary to consider the randomness and ambiguity of investor behavior when modeling the investors’ neighbor herd behavior.These features make the information exchange between investors complex and changeable.Traditional behavioral finance methods and investment theories are difficult to accurately describe the investors’ interaction behaviors.It is more difficult to explain the evolution of information dissemination and emotional diffusion in the investors’ network from a macro perspective.With the rapid development of information technology,investors’ information exchanges have shown rapid proliferation and multi-agent distributed characteristics.Investors’ decisions will be affected by the structure of the information dissemination network.Due to the mutual effects of ambiguity and randomness,fuzzy Bayesian networks is produced.FBN provides a new modeling method for investors’ neighbor herd behavior.It is a useful and necessary supplement to the traditional random diffusion investors’ neighbor herd behavior model.The simulation research of investors’ neighbor herd behavior is based on the FBN model and the neighbor herd behavior theory.This paper starts from the static and dynamic BN model of investor behavior analysis.Then,we construct a simulation model of in fuzzy random environment.The problem of neighbour herd behavior has been widly concerned.First,we analyzes and summarizes the basic theoretical methods of Bayesian networks representation,reasoning and modeling.We discussed the method of applying Bayesian Networks to the simulation modeling of investor herd behavior.We analyzed the advantages of Bayesian Networks in dealing with polymorphism,uncertainty and correlation that prevail in investor sentiment and preferences.It shows that Bayesian network is a powerful tool to solve the problem of investor herd behavior modeling;Second,we analyzed the advantages and disadvantages of the static investor herd behavior Bayesian network.We introduced a time-varying self-learning process and dynamic herding behavior.We propose a dynamic Bayesian network modeling method based on time series and a multi-agent dynamic Bayesian network modeling method.Through the modeling and simulation experiments of the multi-agent herd behavior system in the stock market,we explained the technology,modeling method and derivation process.The simulation system can simulate a variety of stock market environments,proving the effectiveness of dynamic Bayesian networks and providing some references for studying the law of stock market herd behavior.We conduct multiple simulations by building six different types of subjects.Then,We set up different initial market conditions and found differences in LSV herd behavior parameters in different environments.With more hedging,noise and herd investors are shown more herding behavior.However,with more hedging,trending and reverse invsetors are not easy to cause herd behavior.Trending and herding investors show a clear buying herd effect;Finally,we analysis the problem in obtaining conditional probability distribution of multi-parent nodes under fuzzy conditions.Because the investors’ preferences are fuzzy and subjective preferences exist,a multi-agent herd behavior model based on fuzzy Bayesian networks is proposed.This method transforms fuzzy variables into triangular fuzzy numbers based on fuzzy mathematics theory.We use the form of standardized fuzzy numbers to express the conditional probability of investors’ fuzzy psychology of risk preference and the fuzzification process was verified by a calculation example.The simulation model proves that the stock market herd behavior model established by the fuzzy Bayesian network can better fit the fuzzy psychology of investors and has stronger applicability.Similarly,the LSV model is used to measure herd behavior,and the fuzzy Bayesian network provides sampling within a certain interval,which increases the sample space range,simplifies the parameter assignment and calculation process and it can simulate very large samples.Unlike DBN model,the FBN model performs static sampling,so the result is closer to the static Bayesian network.In addition,the Bayesian network designed with Ge Nie software can quickly realize reliability information collection,Bayesian network modeling and reliability analysis.The simulation model developed based on this platform is convenient to use,which is beneficial to the popularization and application of the method.As a typical market irrational behavior,the herd behavior of neighbors will reduce the information transparency,pricing efficiency and forecast accuracy of the financial market.It may lead to financial market turbulence.In addition,a large number of short-term speculation and irrational herding behavior will also cause the speculative habits of funds,weaken the formation of value investment.It is not conducive to the healthy development of the real economy.Therefore,it is of great significance to accurate understand the herding behavior to promote the development of the real economy.It improves the pricing efficiency of the financial market,inhibits the fast in and out of irrational capital and supports the formation of value investment.In a word,based on the investor herd behavior theory and Bayesian network model,this paper constructs a stock market neighbor herd behavior model based on SBN,DBN and FBN to analyze the formation of neighbor herd behavior among investors.Evolution.The model construction process is different from the existing literature research methods,different from the SIR model and the SFI-ASM model.It applies the self-learning and herd learning structure of dynamic Bayesian networks.And it innovatively adds the concept of fuzziness.The triangular fuzzy number method was used to construct the FBN-based stock market herd behavior model and simulated.Based on the theory of investor herd behavior,this paper combines financial theory,Bayesian network theory,investor behavior theory,fuzzy random theory and other theoretical knowledge.We implemented a systematic study of the stock market investor herd behavior and conduct risk measure based on Bayesian Networks.It provides a theoretical basis and simulation model that can be referred to for supervisors to prevent financial risks from spreading among different investors.It has high theoretical value and practical significance. |