The study of resistance to deformation of various steel grades is one of the key issues for the adequate operation of automation systems, which makes it possible to obtain rolled products with the required accuracy in terms of geometric characteristics. In addition, knowledge of deformation resistance is important in the design of rolling mill equipment. In the literature, the values of deformation resistance in the overwhelming majority of cases are given in the form of coefficients of various equations (for example, Hensel-Spittel). However, these formulas often have limitations in the range of technological parameters where they give an acceptable result. It should also be considered that dozens of steel grades are produced at modern rolling mills, and their chemical composition can vary over a wide range depending on the final thickness of rolled products, customer requirements, or based on economic considerations (the most advantageous alloying composition). The study of the rheological properties of such a quantity of materials in the laboratory is expensive, long-term, and labor-intensive, and the literature sources do not provide completeness of the data. The article shows that, using data from industrial rolling mills and machine learning methods, it is possible to obtain information about the rheology of the material with satisfactory accuracy, which makes it possible to avoid laboratory studies. Carrying out such studies is possible due to the high saturation of modern rolling mills with various sensors and measuring instruments. Comparison of the results from industrial data was carried out with the values of the deformation resistance obtained on the Gleeble machine. Based on this comparison, the model was trained based on gradient boosting to take into account the features of the technological process in industrial production.