The cost of comprehensive pipe gallery is higher than that of traditional municipal pipeline facilities, and its investment estimation has the characteristics of having many influence factors, nonlinearity and so on. Considering 10 characteristic factors such as the length of the corridor, the cross-sectional area, the number of cabins, and the number of pipelines, a prediction model was established based on GA-BP neural network, combining with the advantages of Genetic Algorithm (GA) and BP neural network model. Through MATLAB simulation experiments, the investment estimation of comprehensive pipe gallery was predicted and compared with calculating results of traditional BP neural network.The results show that the simulation output value of the test sample is in good linear agreement with the true value of the test sample, and the error is less than 5%. Compared with traditional BP neural network model, GA-BP neural network model has a higher computational accuracy, which indicates that GA-BP neural network model has certain feasibility and effectiveness in engineering applications.