ОVERVIEW OF STATISTICAL METHODS FOR FORECAST DEVELOPMENT

Having a general idea of the nature forecast and understanding methodology forecasting in general, it is advisable for analytical departments, economic management units enterprises, banks, firms (of any socio-economic objects) to develop at least short-term forecasts based on indicators their activity modern conditions coopetitions understand trends changes these indicators. As result conducted research, basic principles that must be observed when developing are revealed, as well detailed description features statistical auto-projective methods given: random walk models containing free term or it; characterizing deterministic trend with fluctuations around trend; moving average models; exponential smoothing using simple smoothing, linear, quadratic seasonal (Brown, Holt, Winters models); integrated presentation autoregressive (parametric ARIMA models). In addition, research confirmed significance developed level confidence obtained future values depends quality models. The criteria checking are: Akaike information criterion, which evaluates model compared each other; Hannan-Quinn Criterion used compare different number parameters an alternative criterion; Schwarz-Bayesian Criterion, compares relative other likelihood function; mean squared error value, absolute value expressed percentage.