Interpretable Modelling of Private Electric Vehicle Adoption in Ireland: Bridging Forecasts and Policy

Abstract

The adoption of electric vehicles (EVs) is a central pathway to reducing road transport emissions and meeting international climate goals. This study develops a two-phase hybrid forecasting and explainability framework to model private EV adoption in Ireland, while accounting for economic conditions, energy prices, policy signals, and public awareness. In the first phase, a Manta Ray Foraging Optimization–Variational Mode Decomposition–Residual Gated Recurrent Unit (MRFO VMD ResGRU) framework is employed to generate multi-horizon forecasts of monthly private EV registrations at 1-, 2-, 4-, and 6-month leads. The VMD-based signal decomposition separates trend, seasonal, and high-frequency components. Each temporal scale is then modeled through the ResGRU architecture, which results in stronger short-, medium-, and long-horizon performance than other models from the same recurrent network family. In the second phase, a surrogate-based explainable learning strategy applies SHapley Additive exPlanations (SHAP) through an Extreme Gradient Boosting (XGBoost) model that mimics the MRFO-VMD-ResGRU forecasts. The explainability analysis identifies annual average wage and the Google Trends Index (GTI), as a proxy for public interest, as the most influential factors associated with private EV adoption, followed by fuel prices. Higher income levels, stronger public interest, and higher petrol prices are positively associated with EV adoption, whereas, interestingly, higher diesel prices exhibit a negative association with uptake. This hybrid time series forecasting strategy with explainable AI gives policymakers and stakeholders transparent, data-driven insights that guide planning for private vehicle electrification and the achievement of long-term emission reduction targets in Ireland.