A spectral prediction model of printer based on GA-BP neural network and subspace partition was proposed in this paper. Color space of printer was divided into subspaces and GA-BP neural network models were applied in subspaces．Spectral reflectance of any printer motivation values can be predicted by GA-BP neural network according to their own subspace. The principal component analysis was used for dimensionality reduction of spectral reflectance, which simplified the neural network structure and maintained the high identification accuracy for the test samples at the same time. Experimental results showed that prediction accuracy of the model improved obviously than the model without subspace partition, which can satisfy the requirement of high-precision spectral prediction of printer.