At present, China's construction cost management lags far behind foreign countries, especially in the study on structural and material characteristics of construction. Based on Radial basis function neural network nonlinear model, the paper selected ordinary commercial civil engineering data of concrete building construction in the 2003-2007 years as samples, chose structural and material characteristics as cost indicators and estimated the cost prices. The results show that the method meets the requirements in the precision. Compared with other methods, it has the advantage of speed and generalization ability. The method provides project managers a better basis for decision making.