Analysis of the possibilities of improving the quality of forecasting prices for certain types of raw materials using the simplest methods of combining individual forecasts

Abstract

Currently, combining forecasts is considered as the easiest way to improve the quality of forecasting compared to individual models.
The purpose of this work is to test the possibilities of the simplest methods of combining, such as a simple average and estimates based on the average squared error of forecasts of previous periods, to improve the qualitative characteristics of short-term forecasts of five indicators of resource prices. The work is based on the forecast base of the Bulletin of Short-term Forecasts of the IEP, which provides an array of primary (combined) forecasts and allows you to build their combinations in real time.
The results obtained allow us to conclude that even the simplest methods of combining can be considered as a way to improve the accuracy of forecasts. In addition, in the case of fluctuations in resource prices, it is even possible to single out a group of methods (namely, combining with weights inversely proportional to the squared errors of individual forecasts) that provide the maximum gain in quality for most periods.

Full document