In recent years, with the rapid development of science and technology, BIM (Building Information Modeling) technology has gradually been widely used. With this technology, various departments in the project can be connected, the cost of the project can be controlled internally, and the project cost can be reasonably managed on this basis. This paper develops the design of the construction cost management system based on BIM and neural network technology, and establishes a cost prediction system based on BIM and BPNN (BP neural network) technology. The BIM-4D technology is used to dynamically simulate the project construction status, form a project database containing the data information of the whole life cycle of the project, and use the BIM automatic quantity calculation function to calculate the physical quantities of the project. Operate in the forecasting system, process the data, and finally get the forecast unit price. Multiply it with the automatically measured quantities to get the corresponding forecast cost information, and complete the dynamic cost forecast. The results show that the relative error between the predicted result and the actual value is less than 8% in 10 groups of data. We can see that the maximum relative error of 10 predicted samples is 3.13%, and the minimum relative error is even 0.65%. Therefore, the prediction effect of the construction project cost prediction model constructed by BPNN is very good, and the relative error is ideal.