A Hybrid Price Prediction Model For Diesel and Gasoline in Ghana
Sampson Agyare *
Department of Mathematical Sciences, University of Mines and Technology, Ghana.
Benjamin Odoi
Department of Mathematical Sciences, University of Mines and Technology, Ghana.
Eric NeeboWiah
Department of Mathematical Sciences, University of Mines and Technology, Ghana.
*Author to whom correspondence should be addressed.
Abstract
The main purpose of this study was to develop a hybrid model that better predict the prices of petrol and diesel in Ghana by using the Box-Junkins autoregressive integrated moving average (ARIMA) and the nonlinear autoregressive neural network-LSTM (NARNN-LSTM) models. The study used a two weekly indicative prices of petrol and diesel from the website of the Bank of Ghana (BoG) and the National Petroleum Authority (NPA) from January 2016 to December 2023. A univariate ARIMA, NARNN-LSTM and a hybrid ARIMA and NARNN-LSTM models were developed and the forecasting performances were compared based on mean absolute error(MAE), mean squared error(MSE) and root mean squared error(RMSE). The ARIMA models performed better than the NARNN-LSTM models for both petrol and diesel based on the error metrics. The hybrid of ARIMA and NARNN-LSTM model far outperformed the base models of ARIMA and the NARNN-LSTM models for MAE, MSE, RMSE for both petrol and diesel in Ghana.
Keywords: ARIMA, NARNN-LSTM, gasoline, diesel, hybrid, forecasting