To predict the shrinkage of molded parts using numerical simulations, the mathematical model should be simplified to overcome the difficulties of formulation due to non-linearity of problems. So it is hard to predict the shrinkage exactly because of the simplification. In the present work, the neural network is used to predict the shrinkage which can implement nonlinear models very well. Comparison between the results of neural network and that of the commercial analysis software, ABAQUS, shows that the result of the neural network is in better agreement with that of the experiments.