The linear regression and neural network methods are adopted to predict the welding compensation quantity for the segmentation construction process of semi-submersible support platform, and the prediction accuracies are compared. It is found that the accuracy of the neural network method is higher than that of the linear regression method. The Root Mean Squared Error (RMSE) of the neural network method is about 0.062 69, which is much smaller than that of the linear regression method. Based on the neural network prediction method, the welding compensation quantity database is designed and built, which provides effective support for high-precision construction of the semi-submersible support platform in the future.