Investment in the petroleum industry is usually faced with a high degree of risk due to uncertainty associated with economic factors. Typical factors include oil and gas price, interest rate, operational and capital expenditure. In addition, the investment risk increases as offshore exploration, drilling and production activities increase. Therefore, accurate prediction of economic factors is crucial in an upstream oil and gas sector in order to make better strategic decisions with minimized risk.
In the present study, four methods of the least square support vector machine (LSSVM), genetic programming (GP), artificial neural network (ANN), and auto-regressive integrated moving average (ARIMA) were initially used to forecast monthly oil price (MOP), daily gas price (DGP), and annual interest rate (AIR). Next, the meta-heuristic bat algorithm (BA) was applied in order to optimally combine the four mentioned forecasting methods in an integrated equation as a novel approach. All required historical data to forecast oil price, gas price and interest rate were collected from the Central Bank of the Islamic Republic of Iran.
Error analysis in terms of coefficient of determination (R2), average absolute relative error percentage (AAREP), root-mean square error (RMSE), and cumulative probability distribution versus absolute relative error percentage were used to compare the prediction performance of forecasting methods.
Error analysis proves that the BA optimized method is superior over all other forecasting methods in terms of highest R2 and lowest RMSE. After the BA optimized method, construction of LSSVM, ARIMA, ANN, and GP has better prediction ability, respectively. The results indicate that the BA optimized method reduces RMSE at least by 6.61% in MOP forecast; by 18.33% in DGP forecast; and by 23.13% in AIR forecast over all other forecasting methods.
