The variation of supplementary components, such as cartridges, has been observed to have an impact on the color electrophotographic (EP) process. The preliminary study shows that the cartridge variation affects the calibration sensor mapping which predicts colorimetric tone reproduction (CTR) values on printing media as a function of internal sensor readings from halftone patterns printed on the printer transfer belt. This paper presents the result of the data-driven study to identify the cartridge patterns and further improve the tone prediction accuracy in calibration of EP platform. First, the constrained hierarchical clustering method was applied to generate a variety of cartridge groups according to the characteristics on the sensor mapping. A new sensor mapping model including a cartridge clustering module was proposed not only to compensate for environmental and consumable conditions, but also to consider cartridge variation. The experimental results show that overall accuracy of the sensor mapping can be improved by ∼ 40% on average after considering the cartridge factor.