Prediction Of Stock Price Based On GARCH Model And BP Neural Network

With the rapid development of the stock market,various experts and scholars have proposed a variety of stock price forecasting methods,mainly including qualitative forecasting methods and traditional quantitative forecasting methods,such as fundamental analysis,technical analysis,moving average,and gray.Model prediction method.But the stock market is a complicated nonlinear dynamic system.The stock price is affected by many external factors,not only including many highly relevant economic,social and political factors,but even the psychological factors of investors will have a huge impact on the stock price.For example,natural disasters and economic crisis have a greater impact on the stock market.Therefore,it is usually very difficult to predict the trend of the stock market.Traditional statistical methods,including moving average,exponential smoothing and linear regression methods have been used to predict the stock price.In addition,the Markov model and the ARMA model have also been used for stock price forecasting.However,the non-linear and time-varying characteristics of market behavior make it difficult for these traditional methods to reveal internal changes in the stock market.This article selects the stock closing prices of ICBC,Golden Sky,Tianbang from January 5,2015 to April 4,2018 as the research object,and describes the stock prices of ICBC,Golden Sky,and Tianbang.The results of statistical analysis,stationarity test,and ARCH analysis have found that ICBC’s stock returns have the characteristics of peak-thick distribution and ARCH effect.Therefore,this paper uses GARCH model and BP neural network model to ICBC’s stock price.prediction.At the same time,due to the timeliness of the stock returns of Jintian High Tech and Tianbang Shares are not very obvious and the ARCH effect is not significant,the ARCH model and BP neural network are not used to fit the stock price.We fit the data to the GARCH model and found that the GARCH(1,1)model is the best from the point of view of model validity and streamlining.BP neural network for ICBC’s stock returns when the number of hidden layer nodes is 5.The degree of fit of the rate data is optimal,and both prediction methods are able to predict the short-term price of the stock.