Abstract

Oil prices movements is very important macroeconomic factor for decision making. The accuracy of results fordifferent types of oil brands depends on models and algorithms. This paper evaluates the effectiveness of usingfuzzy sets to forecast daily Brent oil prices. It also contains possible modifications of the proposed method and incomparison with basic methods. The results suggest that Brent oil prices series have short memory because usinginformation about last 2-days prices shows better forecast accuracy. Forecasting based on fixed universe ofdiscourse shows better efficiency and it also proves that oil prices series has short memory. Adding theprobability of switching between linguistic terms in defuzzification function could be used to improve accuracyof predictions. Also the approach can take into consideration expert’s opinion about direction of future variation.The effective expert’s work can reduce errors of forecast from 1.5% till 0.76%. But this modification can be usedif experts correctly guess the direction of the change in trend in eight out of ten cases and more. The reasonableobtained results can be used by analysts dealing with the prediction of oil prices.

This work is licensed under a Creative Commons Attribution 4.0 License.