An interpretable multi-stage forecasting framework for energy consumption and CO2 emissions for the transportation sector

The transportation sector is deemed one of the primary sources of energy consumption and greenhouse gases throughout the world. To realise and design sustainable transport, it is imperative to comprehend relationships and evaluate interactions among a set of variables, which may influence transport energy consumption and CO2 emissions. Unlike recent published papers, this study strives to achieve a balance between machine learning (ML) model accuracy and model interpretability using the Shapley additive explanation (SHAP) method for forecasting the energy consumption and CO2 emissions in the UK's transportation sector. To this end, this paper proposes an interpretable multi -stage forecasting framework to simultaneously maximise the ML model accuracy and determine the relationship between the predictions and the influential variables by revealing the contribution of each variable to the predictions. For the UK's transportation sector, the experimental results indicate that road carbon intensity is found to be the most contributing variable to both energy consumption and CO2 emissions predictions. Unlike other studies, population and GDP per capita are found to be uninfluential variables. The proposed multi -stage forecasting framework may assist policymakers in making more informed energy decisions and establishing more accurate investment.