How to improve the prediction accuracy of the ANN model to underground water content effectively

In this paper, Hetao plain as a testing ground f and a number of local agro-wells known as the proving prerequisite, use of Induced Polarization (IP) and resistivity sounding and other surface geophysical methods, adding the thickness of the aquifer and the rate of deviation in the relevance of water inflow artificial neural network (ANN) predictive model which based on the parameters of apparent resistivity, polarization rate, half-bad and the decay rate, etc. Then, build a new prediction model of underground water. Through the rules of mean-variance test shows that the accuracy of quantitative prediction of the water content has been greatly improved after adding new input neurons in the exploration area. The study has a good promotion value to the application of ANN forecasting techniques in hydrogeological exploration.