Diesel Fuel Price Prediction Model Based on Multivariate Adaptive Regression Splines Approach

Abstract

Diesel fuel price modelling and prediction is significant to an economy since fuel price has direct consequences on retail commodity prices, transportation and successful implementation of government policies. In this study, an attempt has been made to propose a novel Diesel Fuel Price (DFP) prediction model using Multivariate Adaptive Regression Splines (MARS) approach. The proposed MARS approach was compared with the following benchmark methods: Backpropagation Neural Network (BPNN), Radial Basis Function Neural Network (RBFNN) and Multiple Linear Regression (MLR). The developed prediction models have interest and inflation rates as the input parameters and diesel fuel price (DFP) as the output parameter. A total of 95 data points acquired from Ghana National Petroleum Authority and Bank of Ghana were considered. Of these, 67 served as the training set and 28 was used as testing set. Model validation was done using dimensioned error statistic indicators of Mean Absolute Percentage Error (MAPE), Relative Error Correction (REC) and Noise to Signal Ratio (NSR). Box plot was also used to assess the error distribution of the models’ predictions. Overall, the statistical results revealed that the proposed MARS approach gave the best performance and thus could be used to predict diesel fuel price.