A Comparison Between Hybrid ARIMA-Neural Network Model And GARCH Class Models In Case Study Of Forecasting Stock Market In China

As an important part of capital market, stock market is the barometer of the macro economy, and price index of stock market responds fast to the economy movement. Also, it plays a crucial role in improving capital liquidity, allocating social resources and providing investment opportunities, etc.Having experienced a highly growth over the last20years, stock market in China has reached a considerable scale. By the end of December2012, the overall market capitalization of Shanghai and Shenzhen stock exchanges have reached3697.37billion dollars, ranked as the top three in world’s top stock exchanges.Our stock market has experienced continuous improvement, laws and regulations keep reforming to lead the stock market to a better direction, but there still exist some problems, i.e. liquidity restriction of different classes of stocks, speculation and irrational investor behavior. On the whole the volatility in our stock market is higher than world’s mature capital markets. The abnormal volatility not only makes investors suffer losses, but also destroys the stability of the entire financial system. Therefore it’s very meaningful to study the stock price index in our stock market.Financial time series always present both the linear and non-linear characteristics. This paper introduces a hybrid approach which combines the autoregressive integrated moving average (ARIMA) model and neural networks (NN). First build ARIMA model,neural network, ARIMA-GARCH class models and hybrid model respectively based on the historical closing price of Shanghai composite index and Shenzhen component index from12th, December1996to31th, December2009, then apply them to forecast closing price of Shanghai composite index and Shenzhen component index from4th, January2010to31th, December2012to validation these forecasting models.By comparison the prediction accuracy of different model I get the conclusion that the application of hybrid ARIMA-NN model can improve the accuracy of forecasting. Also, the volatility characteristics of our stock market are obtained from the ARIMA-GARCH models. In the end some suggestions are given to improve the stock market development in China.