Abstract: Industrial injuries nowadays occupy a special place in the production cycle of any enterprise. It is connected first of all with
reliability, stability of the enterprise. Annually, the budget includes the cost of activities related to: improving jobs, staff
development, health, safety, renewal of equipment and inventory. Forecast of industrial injuries risks will give the enterprise an
opportunity for more rational use of the budget. Direction of money resources to those measures which are necessary first of all for
reduction of risks of occurrence of an accident.
In given scientific article the forecast of risks of industrial traumatism occurrence at the enterprise of heavy mechanical
engineering with the purpose of minimization of expenses in the future periods will be made. For this purpose the analysis of
methods of forecasting will be made, in the given research multilayered neural networks of software product Statistica will be
applied. The five best ones have been selected as a method of selection and training. The result of the study will be a forecast for the
next 12 months as well as its analysis. The analysis will be based on statistical data on the number of victims of rolling stock raids in
the branches and farms of JSCo RZD for the period 2005-2015.
Index terms: industrial injuries, forecasting methods, neural networks, time series, exponentiation, training algorithms, forecast, analysis