A building cost estimation model was established based on back propagation (BP) neural networks. The project characteristics determining the cost estimation were obtained from expert experiences. Based on the analysis of 15 typical actual project samples, the BP neural network model subsequently was developed with the EXCEL macrocommands to simulate the nonlinear relationships between the project characters and the costs. The model was applied to the cost estimation of actual projects. The results satisfy the requirements of cost estimation. This new method for cost estimation is efficient and accurate.