Due to the multiple welding parameters and the coupling of various welding parameters in the hybrid welding, any change of the welding parameters will have a significant influence on the weld geometry. Therefore, it is always very important to choose reasonable parameters that can increase the stability of welding bead and geometry. In this paper, we will make the center of welding bead and geometry as the origin of polar coordinate system in which the dimensions of welds will be measured every other 15 degrees. Basing on multiple welding parameters and welding shape dimensions, we apply the BP neural networks to build a calculating model between the welding parameters (welding current, laser power, welding angle, welding gap and welding ) and 24 groups of welding shape dimensions in the polar coordinate system. Finally a genetic optimization algorithm is introduced to improve the forecast precision. The dimension difference between the dimensions on the right side and those on the left side is calculated as the symmetry of the welding shape. The research shows that the dimensions prediction accuracy of the optimized predicting BP neural is 5% and the symmetry prediction accuracy is about 13%. The research method proposed in this paper has a important significance to the hybrid welding research of parameter optimization.