Summary

The existence of objective laws in the world of science, technology makes it possible to compile fairly accurate forecasts of their future development. A class of statistical, or factographic, prediction methods uses different statistical models based on the analysis of quantitative characteristics of forecasting objects, which are the information basis for the development process forecasts. This factual information about the object is a collection of numerical data characterizing the state object prediction at different points in time. The development of methods for recording, describing and analyzing statistical data obtained as a result of observing mass random events constitutes subject of special science - mathematical statistics. Statistical methods of forecasting use the apparatus of mathematical statistics, probability theory and the law of large numbers, which If the sample is large enough and selected randomly, then in it the regularity of the whole population is preserved.

From the point of view of statistical approaches to forecasting this class of methods can be divided into two subclasses. The first is methods forecasting time series, or series of dynamics that allow study the change of the investigated object only in time (one-parameter methods). The second subclass - the methods of multifactor analysis, they allow you to study the change of the object under study both in time and depending on the influence of other factors.

Time series prediction methods include several groups:

- methods of extrapolating trends, which, in turn, include least-squares, exponential smoothing, moving average, S-curve, envelope curves, etc.;

- methods for analyzing publications: analyzing the dynamics of patenting, analysis of the dynamics of scientific and technical publications.

According to a number of researchers, analysis and extrapolation of trends remains the most common approach to forecasting scientific and technological development.

Methods of extrapolation of trends in the development of the investigated object are based on the study of the patterns and tendencies in the development of the object under investigation and their extension to the future, that is, beyond retrospect. At the core methods of extrapolation is the property of inertia inherent in that or other processes and phenomena, especially this is characteristic for large objects of spiders, technology. The bigger object, the greater inertia it possesses, the greater the possibilities of extrapolating the main factors of its development.

Technological development occurs under the influence of a set of constantly (in a certain period of time) acting and random factors. Under the influence of these two groups of factors there is a variation in the levels of the dynamic series. The use of statistical methods makes it possible to distinguish systematic changes from random ones. This kind of trend analysis is called serial ratings. The resulting systematic changes are projected to the future, where time is taken as some artificial measure of all factors that affect on the predicted variable. Modern statistical methods allow allocate systematic linear, logarithmic or exponential changes against the backdrop of unsystematic changes. The regularity of the dynamics of patenting and the publication of information about the state of objects of science, technology in other sources of information is based on methods for analyzing publication flows (patent and non-patent information).