As green iron ore pellets exhibit elastoplastic behaviors, it is crucial to select an appropriate DEM model that accounts for these behaviors to simulate their classification in a roller screen. The most important DEM models include the Hertz-Mindlin and the hysteretic spring. In this study, for the first time, green pellet behaviors were modeled based on their actual shape (multi-spheres) under various conditions of moisture content, bentonite dosage, and iron concentrate type, using the image analysis. All parameters required to simulate the behavior of green pellets were measured through experimental tests. These properties include shape (sphericity), physical (angle of repose), and mechanical characteristics (shear modulus, yield stress, coefficient of restitution as well as static and rolling friction coefficients). Thus, the total contact parameters were adjusted based on these properties. Additionally, a comprehensive comparison was made between the Hertz-Mindlin and hysteretic spring models to accurately simulate the characteristics of the green pellets and assess the efficiency of the roller screen. Finally, it was found that the use of a realistic shape and hysteretic spring model provides a more accurate simulation of pellet classification when applying a roller screen and shows a stronger correlation with the laboratory results.