Proceedings of the 5th World Conference on Climate Change and Global Warming
Year: 2025
DOI:
[PDF]
A Hybrid AI Approach to Comprehensive Carbon Footprint Assessment: Integrating Generative Models, Neural Networks, Knowledge Graphs, and Probabilistic Methods
Sakayong Pattanavekin
ABSTRACT:
Carbon footprint accounting remains a pivotal strategy in the fight against climate change, yet conventional methods often overlook critical Scope 3 emissions due to fragmented data and limited analytical tools. This dissertation introduces a multi-tiered AI framework—encompassing generative models, neural networks, knowledge graphs, and probabilistic approaches—that collectively bolster the accuracy, completeness, and transparency of carbon footprint estimates. Furthermore, it incorporates price as a proxy, excluding luxury goods to avoid skewed outcomes, thereby capturing significant insights about supply chain efficiency. Deployed using real-world data from the Thailand Greenhouse Gas Management Organization, the framework revealed notable improvements over standard approaches, yielding more trustworthy estimates of indirect emissions. Equally important, this research highlights how the synergy of multiple AI paradigms can facilitate a nuanced understanding of carbon-intensive processes, guiding effective mitigation decisions for both policymakers and industry stakeholders. By offering a scalable, evidence-based model tailored to complex emissions scenarios, the dissertation contributes meaningfully to sustainability science and global greenhouse gas reduction initiatives.
keywords: AI Carbon Accounting, Carbon Footprint Assessment by AI, Scope 3 Emissions Assessment