Predicting Nigeria’s petroleum price: A comparison between recurrent neural networks, multilayer perceptron neural networks, and generalized regression neural networks

Abstract

Predicting petroleum pump prices assists in knowing when, how, and what quality to purchase. The purpose of this study is to provide an acceptable deep-learning algorithm for modeling and forecasting Premium Motor Spirits (PMS) pump prices in Nigeria. A deep learning algorithm was employed to capture the non-linear correlations present in the series, detecting patterns and for satisfactory prediction capability. The study utilized 96 monthly petroleum price data from January 2016 to December 2023 as extracted from the National Bureau of Statistics (NBS) websites. The Multifactorial deep learning algorithms like Recurrent Neural Networks (RNN), Multilayer Perceptron Neural Networks (MLP-NN), and Generalized Regression Neural Networks (GR-NN) were employed and compared to determine the best model that provides the most accurate method for prediction. Results generated from predicting evaluation criteria like Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) suggested that the Multi-layer Perceptron Neural Network model outperforms the Recurrent Neural Networks and Generalized Regression Neural Network models in the out-of-sample prediction performances of the models. The prediction using the GR-NN model revealed a relative increase in the PMS prices. The results further projected PMS Price as high as N727 per liter at the end of 2024 and above N2,213 for the year 2025. The study therefore recommends that Multilayer Perceptron Neural Networks provide an optimal deep learning algorithm for modeling and predicting Nigeria’s PMS prices.