Process charcteristics and prediction of weld shapes were studied for spot welding of thin plate using Nd-YAG laser. Weld variables such as plate thckness, gap size, focal length, pulse energy, pulse duration were selected and varied to make various experiments. Finit difference method and neural network theory were applied to predict nugget size and penetration depth for weld shapes. In case of specimens with no gap between thin plates, finit defference method was used to calculate penetration depth and nugget size. The calculated results were compared with experimental results to verify the proposed finite difference method. Neural network theory was applied for specimens with gap between thin plate. Three different combinations of weld variables were considered to find out the appropriate learning results. The reliability of proposed model for neural network was verified in comparison with error percentage between experimental data and estimated data. In application, combined model of finite difference method and neural network was applied for the prediction of bead shapes. Finally evaluation program based on learning results to neural network was developed to predict the configuration of weld shapes for various welding conditions which were not experimented.