Formation of a Credible Methodological Framework for Epidemiological Forecasting of the Incidence of Hemorrhagic Fever with Renal Syndrome Using Machine Learning Techniques

The high medical and social significance of hemorrhagic fever with renal syndrome (HFRS) in Russia is a valid argument in favor of developing qualitatively new approaches to predicting epidemiological complications. The complex nature of interaction of factors determining the intensity of epidemic manifestations requires the continuation of the search for the most reliable prognostic criteria for pre-epidemic diagnosis of HFRS, providing the completeness of information that is necessary in modern conditions to optimize the epidemic process management. The aim of the work was to form an objective methodological basis for predicting the epidemiological situation on HFRS through the use of machine learning in modeling the effect of combinations of abiotic risk factors on the actual incidence of HFRS. Materials and methods. The study materials comprised the data on 10,788 cases of HFRS, registered in the Saratov Region between 1982 and 2022. More than 46 thousand values of meteorological indicators obtained from the archive of the database of meteorological observations for the period of 1981–2023 of the Saratov – South-East meteorological station, presented in the open access by the Hydrometeorological Center of Russia, are considered as factors likely to influence the epidemic process of HFRS. The development of a neural network forecasting model is based on a specialized module of the Loginom platform version 7.1. Results and discussion. The paper presents the advantages of neural network forecasting of complications of the epidemiological situation on HFRS based on the analysis of complex, nonlinear dependencies of the influence of abiotic factors on the morbidity among the population. A standard scenario for neural network forecasting of epidemic complications due to HFRS has been developed and a prognostic model of incidence in the Saratov Region has been tested. It is shown that the use of neural network algorithms allows to obtain a forecast of the possible development of the epidemiological situation on HFRS in the future with a probability of 98.8 %, which will ensure a qualitative transition from expert forecasting to independent analysis of epidemiological trends, significantly increasing the information capacities and importance of forecasting when planning and conducting preventive measures by the efforts of Rospotrebnadzor institutions.