Abstract: To explore the relationship between process parameters and joint performance in magnetic field-assisted magnesium/aluminum laser welding and to establish a predictive model for guiding process parameter design, an experimental design method was adopted. Laser power, welding speed, and magnetic field intensity were selected as variables to investigate their effects on welding joint performance. A predictive model for welding joints was developed based on the random forest (RF) algorithm, and the dung beetle optimization (DBO) was employed to optimize the model's key parameters (number of trees and number of leaves). The results demonstrate that joint performance is better when the weld morphology coefficient ranges between 1.37 and 1.58. The relative importance of laser power, welding speed, and magnetic field intensity on joint performance is 0.608, 0.212, and 0.276, respectively. The coefficient of determination R² on the test set was improved from 0.742 to 0.950 by the DBO-optimized Random Forest model (DBO-RF), with significant enhancements observed in the model's generalization ability, overall accuracy, and computational speed. These findings provide a basis for the design of process parameters in magnetic field-assisted laser welding.