Predicting petrol prices in Zambia using machine learning and explainable artificial intelligence under a regulated cost-plus pricing framework

Abstract

Fuel price volatility presents significant economic and policy challenges in import-dependent economies such as Zambia. This study implements and compares four machine learning models: Ridge Linear Regression (LR), Random Forest (RF), Support Vector Regression (SVR) and Long Short-Term Memory (LSTM), using the Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Coefficient of Determination (R2) and the Diebold–Mariano (DM) statistic as the performance evaluation metrics for petrol price prediction. The dataset used in this research consisted of the petrol prices as the dependent variable and the International Oil Price (IOP), Zambian Kwacha to United States Dollar exchange rate (ZMW/US$), excise duty, Value Added Tax (VAT), and inflation as the independent variables, covering a period from January 2011 to September 2025. The modelling framework was evaluated within Zambia’s regulated cost-plus fuel pricing system, where petrol prices reflect both market conditions and administratively determined fiscal components. The performance evaluation revealed that Ridge Regression consistently outperformed the competing models, achieving a test MAE of 0.1302, RMSE of 0.3783, and R2 of 0.9814. Explainable Artificial Intelligence (XAI) techniques, including Shapley Additive Explanations (SHAP) values, feature importance, and partial dependence plots (PDPs), were integrated to enhance interpretability. The XAI results indicate that the IOP and the ZMW/US$ exchange rate are the dominant drivers of petrol price predictions, while inflation exhibits a comparatively weaker influence and fiscal variables contribute limited prediction-level variation.