Ecological Systems and Devices Annotation << Back Flue Gas Temperature Prediction WhenOperating Burner Devices Kuvayskova Yu.E., Kovalnogov V.N., Klyachkin V.N. Based on the known results of operation and testing of the burner device, it is necessary to build a mathematical model to predict the temperature of flue gases depending on the value of performance indicators (load, air flow, methane and biogas, fuel and oxidizer compositions, etc.). This is the task of constructing a regression dependence, but standard methods of regression analysis often do not provide the necessary accuracy of forecasting. In this case, special methods of machine learning based on precedents – "with a supervision" – may be more effective: compositional methods – random forest or gradient boosting, support vector method, and others. At the same time, the result of the forecast significantly depends on the hyperparameters of the method. A modified grid method is proposed to optimize the selection of hyperparameters. The purpose of the study is to develop an algorithm for building a mathematical model that provides the necessary accuracy of forecasting. It is shown that in the example under consideration, the support vector method showed the best accuracy. Keywords: regression model, machine learning, decision tree busting, random forest, support vector method. DOI: 10.25791/esip.4.2025.1516 Pp. 45-50.