No abstract

Epizootical monitoring of natural tularemia focuses for 1976-4014 years showed that main carrier of infection is common vole (Microtus arvalis). For the part of all obtained cultures of tularemia bacteria common vole has 51,2%. Its strength and infection can change. At period for 1976-2006 years, natural tularemia focuses were at high activity level, but since 2007 year for now, they are at low activity level. By using method artificial neural network made mathematical model of prediction epizootical tularemia; found that climatic factors (downfalls, temperature, snow depth) playing main role at processes reduction activity and strength of natural tularemia focuses, for example, tularemia epizootia can happen in case of froze December, numerous downfalls and high snow depth.

Regression analysis - a set of statistical methods for processing of experimental data to a condition of stochastic dependence study of the value of non-random or random variables to define this relationship. Statement of the problem of regression analysis is formulated as follows. There is a set of observational results. Requires a quantitative relationship between the index and factors. In this paper, we try to establish a quantitative relationship between the incidence of natural - focal infections and biotic and abiotic factors of the environment. By biotic factors include: the number of infection and the major carriers and vectors to abiotic factors - weather (average monthly air temperature, the monthly average rainfall, snow depth in December, January, February, March). When studying the effect of 22 factors on the incidence of leptospirosis using multiple regression the mathematical model, which has a low level of trust, and when using the stepwise regression established the influence of one factor - infection of the common vole of the 22 factors. Level of trust models and model coefficients are significant. This method allows to determine only the linear relationship between the incidence and natural factors, as in the case of the nonlinear coupling tightness does not install. Natural foci of infection is a complex ecological system. Based on the terms of modeling complex systems, which may include: the possible impact of non-linear elements in the output parameter, synergy and reciprocity under the joint influence of individual factors, the need to address in some cases categorical factors and multiple output parameters of a complex system, it is necessary to choose an artificial neural network (ANN), allowing to realize these conditions in the preparation of a mathematical model of the system.

Mathematical methods and models used in forecasting problems may relate to a wide variety of topics: from the regression analysis, time series analysis, formulation and evaluation of expert opinions, simulation, systems of simultaneous equations, discriminant analysis, logit and probit models, logical unit decision functions, variance or covariance analysis, rank correlation and contingency tables, etc. In the analysis of the phenomenon over a long timeperiod, for example, the incidence of long-term dynamics with a forecast of further development of the process, you should use the time series, which is influenced by the following factors: • Emerging trends of the series (the trend in cumulative long-term effects of many factors on the dynamics of the phenomenon under study - ascending or descending); • forming a series of cyclical fluctuations related to the seasonality of the disease; • random factors. In our study, we conducted a study to identify cyclical time series of long-term dynamics of morbidity of HFRS and autumn bank vole population. This study was performed using the autocorrelation coefficient. As a result of time-series studies of incidence of HFRS, indicators autumn bank vole population revealed no recurrence, and these figures are random variables, which is confirmed by three tests: nonrepeatability of time series, the assessment increase and decrease time-series analysis of the sum of squares. This shows that a number of indicators of the time series are random variables, contains a strong non-linear trend, to identify which need further analysis, for example by means of regression analysis.