In view of the frontier scientific problem that the cost of the prefabricated building is higher than the traditional building, and prefabricated building is easy to out of control of the investment. According to the related literature and the related characteristics of the prefabricated architecture, the characteristic parameters of the investment estimation for prefabricated building by using the significance characteristic theory are obtained. The related characteristic parameters are introduced into the BP neural network model as the input vector, repeated training and the established the estimated model. The prefabricated project of a primary school in Changsha is taken as a sample to test the feasibility of the model. The results show that there is a high degree of coincidence between the simulated output value of the sample, and the result error is within 3%. The accuracy of the result can reach the requirement standard of the investment estimation error in the early decision stage of the prefabricated building project. It has practical feasibility and can be used for the investment of the prefabricated building estimates.