Abstract: A multi-factor weighting method was used to analyze the parameters influencing weld bead geometry of hyperbaric GMAW. The orthogonal test results of hyperbaric GMAW were taken as training samples and the prediction model for weld bead geometry was established via BP neural network. Accuracy of the model was verified by comparing the prediction results with the actual welding test results. Based on the weights and thresholds in the prediction model, the weighting coefficient of each factor to the parameters of weld bead geometry was obtained by the Garson algorithm, which was verified by experiments. The results show that the weighting coefficients obtained by the multi-factor weighting method were consistent with the variation tendencies of the parameters of weld bead geometry with various factors and the weighting coefficients had a guiding role for the engineering application of hyperbaric GMAW.