出版社：Canadian Research & Development Center of Sciences and Cultures
摘要：Forecasting the stock market price index is a challenging task. Many scholars have tried on many kinds of models to predict the stock index, mainly autoregressive integrated moving average model (ARIMA), artificial neural networks (ANN) with genetic algorithms (GA). This paper documents a set of thorough empirical tests of ANN's with different choices of inputs and different numbers of hidden neurons for forecasting the CSI 300 - the benchmark stock index of China. The prediction accuracy is measured in terms of hit rate and mean square error. The trend of the hit rate is observed by adjusting the window length and the number of hidden neurons. The results show that the hit rate is highest when the window length is between 14 days to 20 days.
其他摘要：Forecasting the stock market price index is a challenging task. Many scholars have tried on many kinds of models to predict the stock index, mainly autoregressive integrated moving average model (ARIMA), artificial neural networks (ANN) with genetic algorithms (GA). This paper documents a set of thorough empirical tests of ANN's with different choices of inputs and different numbers of hidden neurons for forecasting the CSI 300 - the benchmark stock index of China. The prediction accuracy is measured in terms of hit rate and mean square error. The trend of the hit rate is observed by adjusting the window length and the number of hidden neurons. The results show that the hit rate is highest when the window length is between 14 days to 20 days.
