Using the mass ratio of ethylene-vinyl acetate to EVA, glycerol content and NaHCO3 content as the input parameters, the viscosity as the output parameters, a 3-layer BP (back propagation) neural network was established. The melt viscosity of composite foaming material was tested by capillary rheometer, while the results were taken as samples to forecast the properties of starch foaming materials. The results showed that the BP neural network could predict the properties with fairly good accuracy. Meanwhile, the viscosity of foaming material increased with the increase of EVA content, the viscosity of foaming material decreased with the increase of glycerol content and NaHCO3 content.