It is important to select input variables when the neural network forecasting model is proposed. In this pa-per, by using the autocorrelation function on input
In this paper we propose an experimental method to choose a prior distribution. Different from many re-searchers, who offered lots of principles that separated