In order to improve the accuracy of the assembly man hour prediction of ship intermediate products, a two-stage man hour prediction model based on big data and considering error correction is proposed on the basis of considering the assembly process characteristics and the influence of environment, personnel and other factors on the man hour prediction. In the first stage, the process information of intermediate product assembly is extracted from the ship design software, and the BP neural network model is established to preliminarily predict the working hours; In the second stage, the error correction model of working hour prediction based on xgboost algorithm is established by collecting big data of external factors that may affect working hour prediction; Finally, add the two-stage prediction results to obtain the man hour prediction value of the job. The validity of the model is verified by the calculation of an example, and it can provide a practical solution for the prediction of working hours of general machining products and the research of the direction of working hours correction.