Abstract: Aiming at the difficulty in precise control of weld bead geometry for low carbon steel fabricated by cold metal transfer wire arc additive manufacturing, ER120S-G low carbon steel welding wire was adopted as the experimental material. A four-factor four-level orthogonal test was performed for process experiments. Range analysis and variance analysis were combined to systematically investigate the influence priority and variation laws of wire feeding speed, welding speed, dry extension and protective gas flow rate on bead width, reinforcement and cross sectional area. Based on the orthogonal test dataset, a support vector regression (SVR) prediction model integrated with Bayesian optimization and 10-fold cross-validation was constructed. The generalization performance of various kernel functions was compared, and the quadratic polynomial kernel was selected to establish the prediction model for weld bead morphology. The results show that wire feeding speed and welding speed exert significantly greater influences on weld bead geometry than dry extension and protective gas flow rate. Increasing wire feeding speed leads to larger bead dimensions, whereas increasing welding speed reduces bead dimensions. Subject to the coupling effect of resistance heat and heat dissipation for dry extension, the bead dimensions present a variation characteristic of decreasing first and then increasing. The coefficients of determination R² of the established SVR model for bead width, reinforcement and cross sectional area reach 0.965, 0.962 and 0.936, respectively. High-precision prediction of bead forming dimensions under small-sample conditions is realized, which can provide theoretical and data support for the forming quality control of low-carbon steel components fabricated by cold metal transfer wire arc additive manufacturing.