This study deals with the laser cutting of stencil for the PCB. The most important aim of this study is to determine optimal conditions which make good-qualify stencil in Nd:YAG laser cutting. We made an experiment according to various variables (power. type of mask. gas pressure, cutting speed, and pulse width) and analyzed the cutting characteristics (surface roughness, kerf width. dross) . Each variable has optimal value for good-qualify cut edge under fixed condition. And neural network after learning experimental data with a million time iteration could predict surface roughness of cut edge under arbitrary condition approximately.