The work aims to establish a spectral prediction model of spot color overprint color based on the relationship between density and spectral reflectance. Firstly, the relationship between front and back scene density and spectral reflectance was built according to the density definition formula. Based on this, a calculation method of spectral reflectance of overprint color was constructed. Secondly, the model was optimized, and the spectral prediction model of spot color overprint color was established. 30 spot color overprint color samples were printed by gravure. Based on Malab, the prediction model was verified with color difference and fitting coefficient R2. Finally, another batch of spot color overprint color samples were printed under the same process conditions, and the accuracy of the model was checked by color difference. The experimental results showed that under the same dot area rate, there was a strong correlation between different spot color overprint color and its corresponding front and back scene density, and the fitting coefficient R2 was greater than 0.9. In the accuracy test of the model, the color difference between the predicted value of the model and the measured value of the sample was between 0 and 2.5 CIE1976 for 73.7%, and the color difference between the majority of the color blocks was between 0 and 2 CIE1976, accounting for 60.54% of the total frequency. The overall average color difference was 1.92 CIE1976. Therefore, the model proposed has high prediction accuracy and can basically meet the demand for color prediction