Neural Networks Based on Zero Cost Function for Stock Price Researching
Weixing Wei · Journal of Guangxi University for Nationalities · 2009
First of all,this paper selects 40 days closing price of Shanghai's stock index in 2004 and 2007,which is used as training sample.We make use of zero-cost function neural network algorithm to train and simulate for a three-layered neural network.It's successful to implement precise I/O mapping,the error is almost zero in this simulation.Secondly,the data in 2007 and 2006 is as training sample and test sample separately,which is used to test generalization ability of network.