Aimed at uncertainty of the machining mechanism of wire electrical discharge machining (WEDM), the relationship between process parameters and process indexes is highly nonlinear. Taguchi experiment was designed to analyze the impact of pulseon time, pulseoff time, pulse current, wire speed and tracking coefficient on the material removal rate (MRR) of during machining SKD11. Generalized regression model (GRM) and Grey forecasting model (GFM) were respectively designed, and the two models were proved by validation experiments and compared to seek the optimal process parameters. Validation experiments result shows that GFM which is built can accurately predict the MRR in actual machining.