ABSTRACT
The article presents original multi-factorial linear regression models developed on the basis of panel structure of data. The distribution of the regions of the Russian Federation according tuberculosis burden was implemented using cluster analysis technique. The evaluation of degree of impact of analyzed factors to epidemic process was implemented using Pratt metric. In overall, approximately 100,000 records from 78 regions of the Russian Federation were processed. four regional clusters were marked out differentiating by level of morbidity, prevalence and mortality of tuberculosis. The diversity of social economic factors impacting the main indices of epidemic process is reducing as its intensity increases. If in the first cluster 7 factors impacted on morbidity and 4 factors -- on mortality, then in the fourth cluster only 1 factor impacted morbidity and 2 -- on mortality. At that, intensity of their impact on epidemic process increased significantly. So, for example, the role of such factor as "Area of housing per capita" in decreasing of tuberculosis morbidity among regions of the 4th cluster assessed in 79%, and among regions of 1st cluster only in 17%. The study results demonstrated that modern techniques of machine learning permit to develop models for quantification assessing of impact of social economic conditions of a particular region on activity of epidemic process of tuberculosis.