Prediction of oil price variations with ARIMA optimization model, innovating with gross operating force

Abstract

The present study evaluates the effectiveness of the multivariable ARIMA model with brute force for the case of the oil price, predicting the behavior of the shares in the following week of a last analyzed date. The objective is to construct a predictive model with a percentage of prediction higher than 50% and, therefore, to improve the decision making for the investors. We used the available information on the oil quotation and shares of the financial web site of three companies, Exxon Mobil, Gazprom and Rosneft, during the period from February 4th, 2011, to February 4th, 2016. It was possible to observe the variation of prices, and to compare the actual data with the variations predicted with the model. We used 12 variables, generating 100,000 random iterations with brute force, without simplex and/or solver optimization, which limited the obtaining results. With the brute-force technique, a prediction capacity of more than 60% could be established for the case of oil prices and oil company stocks.