This paper presents bootstrap algorithm, that helps increase the accuracy of the forecasts creates with additive Holt-Winters model for seasonal times series. In the process of forecasting was used the overlapping blocks bootstrap method (Kunsh method) with various blocks lengths. In calculates was used "forecast" and "tseries" modules from statistical program R.
The following study presents the empirical analysis of the numeric methods in forecasting in conditions of lack of full information. In forecasting the following methods were used: segment, two variants of curves methods, and four variants of Lagrange methods. In analysis are used the average relative forecast errors in six variants of blanks. This study is an attempt to answer a question, whether the amount and distribution of blanks affect the quality of forecasts.