Aiming at the problems of lag and imprecision in off-line color detection of printed material, a color prediction model of liquid water-based ink based on near-infrared spectroscopy was proposed. Multivariate Scatter Correction (MSC), Standard Normal Variate (SNV) and Savitzky-Golay filter (SG) were used to preprocess the original spectral data. Partial Least Squares Regression (PLSR) and Principal Component Regression (PCR) prediction models were established for the original spectral data and the preprocessed spectral data with the Lab value of the printed material respectively. The results show that the prediction accuracy of PLSR prediction model based on MSC preprocessing is the highest, R2 of L, a and b are up to 0.9885, 0.9879 and 0.9938 respectively, and the average color difference of predicted colors is 0.71. Near infrared spectroscopy of liquid water-based ink can accurately predict the color of printed material, which provides a new idea for online detection of printed material.