This article proposes a method for a comprehensive analysis of the controlled parameters of the technological process and the dynamics of sinter cake disintegration during mechanical processing, taking into account the transport lag of the measured output characteristics (fines yield 0-5 mm when sinter cake is destroyed within the sinter plant and along the conveyor path for transporting sinter to the blast furnace shop) from the current composition of the charge and the parameters of the sintering process. On the basis of the specified output characteristics, an assessment of the sinter yield and strength was carried out. To analyze and predict the physical and mechanical properties of iron ore sinter, a new method of constructing regression equations based on linear programming problems with partially Boolean variables was used. The most significant drawback of the standard approach is the strong dependence of the regression equation coefficients on random anomalous observations (outliers), which in practical problems can be quite large and it is impossible to filter them out. The proposed MILP Regression method improves the forecast quality by automatically excluding atypical observations from consideration. The method is compared with other regression models. The method of mixed integer linear programming (MLIP) was used to process the current information about the operation of the sinter shop No. 3. The most significant parameters that affect the formation of the total amount of fines during the destruction of sinter were established: the component composition of the charge, sinter machine pallets speed, pressure gas and temperature in collectors, temperature in ignition furnaces and behind them, temperature in last vacuum chambers of sintering machines. The obtained regression equations quite accurately reflect the influence of the initial data array on the yield of recycled sintering products and they can be used to predict the output parameters of the sinter quality. A. A. Eliseev, manager on raw materials research (PJSC Severstal), Yu. A. Malygin, director (JSC Ferroks), N. P. Chernavin, assistant of the department “Big data analytics and video analysis methods” (Ural federal university), F. P. Chernavin, associate professor, department “Simulation of managing systems” (Ural federal university) participated in preparation of this article.