Recently, the engineering designer of injection mold has become more and more dependent on the CAE. In the design factors of injection mold, the shrinkage rate should be considered as one of the important performances to produce the reliable products. Therefore, the shrinkage rate can be mostly calculated by the MoldFlow and CATIA in the design process. However, it is not easy to predict the shrinkage rate of a plastic injection mold in its design process because the analysis can take minutes to hours, the high computational costs of performing the analysis limit their use in design optimization. In this study, the surrogate models based on the Taguchi Design in order to optimize the shrinkage rate of bevel gear injection mold is used instead of the original models, facilitating design optimal design.