AI-ASSISTED PREDICTIVE ANALYTICS FOR ROAD-TO-RAIL FREIGHT MODAL SHIFT AND TRANSPORT DECARBONIZATION

Abstract

Pakistan’s transport sector faces mounting sustainability challenges due to rapid motorization and a freight system dominated by aging, diesel-powered road vehicles. Although motorcycles account for more than 80% of the national vehicle fleet, heavy-duty freight vehicles contribute nearly 60% of transport-sector CO₂ emissions, indicating a fundamental structural inefficiency. This study presents a hybrid predictive analytics framework combining an ARIMA-based time-series model with machine learning techniques to forecast transport emissions and evaluate road-to-rail freight modal shift scenarios. Using national fleet composition, fuel consumption, and emissions data for financial year 2023–24, ARIMA is employed to establish a business-as-usual emissions trajectory, while supervised machine learning models capture nonlinear relationships between freight activity, fuel use, and modal share. Multiple modal shift scenarios are simulated to quantify the emissions and energy impacts of increased rail freight penetration. The results demonstrate that even moderate shifts from road to rail can yield substantial reductions in carbon emissions and fuel demand. The proposed AI-assisted framework provides a data-driven decision-support tool for transport policy planning, supporting climate-resilient and energy-efficient freight systems in emerging economies.