JUXTAPOSITION OF SHORT-TERM PROGNOSTICATION METHODS FOR SEASONAL INDUSTRIAL PROCESSES

Abstract

The present work represents investigation into the issue of short-term forecasting and prognostication methods, along with their juxtaposition. The calculations are based on figures of a firm distinguished by its seasonal production process. This means that the fi rm’s income function has seasonal character in time measurement. The expenditure function that is basic for the research conducted appears seasonal as well, but at a slower tempo than the income function. As a rule, optimization of fi rms’ management at large and individual operational periods planning both rest on appliance of econometric modeling and prognostication of changes in industrial processes’ figures. Forecast always turns multichoice and thus suggesting several alternative solutions. This is why multiplicative econometric model of time series dynamics as well as proportional predictive model for one period in advance is being formed on the basis of aforementioned figures. Hereupon, errors of models’ calculation are being computed and compared. Th e essential condition for conduct of this research proves to be the availability of a low number of time series observations data (up to 12), which obstructs precision of calculations in each case under consideration. The research held has demonstrated that the error rate for a model formed via proportional method of prognostication recedes the one of regression autocorrelation model, with the mean relative error’s value signifi cantly shift ing downwards as well. In the meanwhile, it is the regression autocorrelation model that better reflects the cumulative variation of expenditure function’s time series levels for preceding periods. In this case a model’s quality rating also turns of rather high predictive value. Both models imply calculations of expenditures predictive values that logically fit with existing time series, whereas the proportional model lends higher tempo of the rate’s increase. Hereby it appears expedient to resort to both methods of expenditure amounts prognostication for one operational period in advance. Furthermore, the research has proved that the proportional method holds a certain advantage, since here prognostication requires only two discrete quantities being expenditure values of two preceding periods. At the same time, the multiplicative model requires sequential series of points marking expenditure values for periods at advisable number of over 12.