Artificial Intelligence-Based Forecasting of Vehicle Carbon

Dioxide Emissions A Review of Machine Learning Models,

Optimization Strategies, and Sustainable Transportation

Applications

Mahmoud Elshabrawy Mohamed1,*

1 Computer Science and Intelligent Systems Research Center, Blacksburg 24060, Virginia, USA

Received: November 23, 2024 Revised: December 27, 2024 Accepted: January 17, 2025 ⋆ Corresponding author

ABSTRACT

Forecasting carbon dioxide emissions from vehicles is essential for sustainable transportation, climate-aware urban

planning, and intelligent environmental management, as road transport contributes substantially to energy-related

greenhouse gas emissions. Rapid growth in vehicle ownership, mobility demand, fuel use, freight activity, and

congestion has exposed the limits of conventional emission-estimation methods, particularly when nonlinear and

multi-factor relationships must be represented. This review examines how artificial intelligence and machine

learning support vehicle carbon dioxide emission forecasting through data-driven models, deep learning, feature

selection, hyperparameter optimization, and metaheuristic algorithms. It emphasizes the value of integrating traffic

volume, vehicle attributes, fuel type, speed patterns, road conditions, temporal behavior, weather variables, and

energy-consumption indicators to improve prediction reliability under real-world conditions. Optimized forecasting

models can support emission monitoring, mitigation planning, traffic management, and low-carbon mobility policies,

although challenges remain in data availability, interpretability, uncertainty, regional transferability, computational

cost, and transparent validation.

Keywords: Vehicle carbon dioxide emissions; Emission forecasting; Artificial intelligence; Machine learning; Sustainable

transportation.

1. INTRODUCTION

The forecasting of vehicle-related CO2 emissions has become

a central research direction in sustainable transportation, environmental

intelligence, and low-carbon urban governance

because transport systems increasingly operate under the combined

pressure of mobility growth, energy demand, congestion,

freight expansion, and climate-mitigation obligations.

Vehicle emissions are not produced by a single isolated factor;

rather, they emerge from the interaction between traffic flow,

vehicle type, fuel or energy source, route selection, driving

speed, operating conditions, road geometry, logistics demand,

and broader socio-economic activity. This complexity makes

vehicle CO2 forecasting fundamentally different from conventional

static emission accounting, since prediction models

must capture temporal variation, nonlinear relationships, and

region-specific mobility behavior. Within this context, artificial

intelligence-based forecasting provides a promising analytical

foundation because it can learn hidden dependencies

from heterogeneous data sources and transform raw transportation

indicators into actionable emission estimates. Re-