The aim of this study was to establish a printer spectral prediction model using RBF (Radial Basis Function) neural network based on subspace partition. The color space of printer was divided into subspaces and RBF neural network models were applied in subspaces with least square method． Spectral reflectance of any printer motivation values were predicted by RBF neural network according to their own space. Experimental results showed that prediction accuracy of the model was obviously improved compared with models without subspace partition, which can satisfy the requirement of high-precision spectral prediction of printer.