Analysis Of The Herding Effect Of China's Stock Market Based On QRNN Model

In the face of a decision,a person's final choice is often influenced by many factors around him,and the impact of human beings is particularly important.People are always inadvertently learn from their predecessors choice,this herding behavior also applies to financial markets.Investors will always wait and see the direction of the market before investing,or listen to the advice of professional investors,so it is important to analyze the herding behavior of investors in financial market decisions.Based on the construction of CSAD index,this paper uses the quantile regression neural network method to analyze the stock market in China.First,the neural network can use the median or other quantile as the dependent variable in the established regression model,and overcome the characteristics that the traditional regression can only use the mean as the dependent variable.Secondly,the neural network can simulate and estimate the possible non-linear structure,without the need to specify the exact form of the function,so that the maximum possible to fit out the real curve trend.The fractional regression overcomes the limitation that the mean regression can only reveal the influence of the explanatory variables on the mean value of the explanatory variables,and can show the effect of the explanatory variables on the whole distribution of the explanatory variables,so that more detailed information can be obtained.The results of the analysis of the herding behavior of the same research object in different periods were observed to observe the degree of maturation of the mainland stock market,and the results were analyzed by the regression of the quantile regression.Point of the stock market changes in the effectiveness of herding.Then,the "horizontal" stratification analysis of the research object was carried out to study the change of the herding behavior of different enterprises in different stock positions at different times and different points in the same stock market.On the basis of summarizing the different regression methods,we use the quantile regression neural network method to analyze the A stock index of the above stock market and the yield of each included stock.Based on the "vertical"stratification analysis of the same index data,the performance of the same research object in different periods was compared with the results of the regression.The results of the quantile regression were used to analyze the different nature of the enterprises the change of herding behavior in stock market at different locations.The results show that:(1)the correlation between CSAD and market rate is very significant,and the flock effect is obvious at most points in two periods.In the bull market and the bear market,the decision-making behavior is different,and with the sub-site movement there is a "cross" phenomenon,low score at the flock effect significantly reduced "obvious",high score points The significance of herding has been enhanced.(2)In the two-period comparative analysis,the performance of private enterprises was significantly better than the data of state-owned enterprises,and after the financial crisis and large-scale rescue policy in the local low score points private enterprise flock effect is no longer Significantly.(3)In the stock market,investors in private equity firms are more "independent" than investors in state-owned enterprises and are more responsive to changes in the stock market.Finally,the corresponding suggestions and measures are given for the corresponding conclusions.