Ecological Systems and Devices Annotation << Back Development of a Model for Forecasting Forest Fires Based on Climatic Data Kolmogorova S.S., Ivanov S.A., Larionov A.D. The article proposes an approach to forest fire forecasting based on a model trained on meteorological data. Binary classification methods are considered, analyzed and defined for use. The logistic regression method was chosen to train the model when solving the current problem. Its main advantage over others is the ability to work with heterogeneous input data, as well as to take into account restrictions on probability values, due to which logistic regression finds its application in finding transition probabilities of the system. The data for model training were collected in the form of tabular values and converted to xlsx format. The period to which the model refers during training is 4 years, taking into account the seasonality of fi res in the Tyumen region: April, May, June, July, August, September. The result of the study was the development and training of a model for forecasting forest fires based on weather parameters with a relevant sample of data from open sources, a software solution was implemented and the forecasting efficiency was verified. Keywords: binary classification, predictive model, forest fires. DOI: 10.25791/esip.11.2025.1554 Pp. 13-21.