A Data-Driven Framework for Estimating Hard-to-Measure Value-Chain Emissions

Budhraja, Lavanya (2025) A Data-Driven Framework for Estimating Hard-to-Measure Value-Chain Emissions. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

The precise measurement of corporate carbon footprints, especially indirect Scope 3 emissions, has become one of the urgent issues facing business enterprises with the intention of addressing their sustainability goals. The existing conventional carbon accounting approaches are associated with the problem of information fragmentation and enormous truncation errors, which restrict the possibilities of the organizations to state the overall effects on the environment. This study fills these gaps by creating a machine-enhanced automated and data-driven structure which can predict emissions using primary operation-related data. The research was carried out through a strict and multistage analytical pipeline made to suit the complexities of the environmental data. To look at the technical feasibility of the proposed solution, an artificial sample of 23,808 corporate profiles was used. This domain-specific imputation was used to determine the quality of data, K-Means clustering (k=4) to realize strategic segmentation, and a set of ensemble predictive models to obtain quantitative emission. The experimental results showed that the XGBoost regressor model performed best in the prediction of aggregate emissions with an R² of 0.52, but a Multi-Output Random Forest model performed better in the direct Scope 1 emissions (R² of 0.54) than in the stochastic Scope 3 emissions (R² of 0.47). The structural analysis revealed existence of specific operational clusters, which signified there was need to have sector-specific baseline so as to have proper reporting. The paper presents a scalable and replicable artefact that demonstrates the viability of algorithmic carbon accounting and provides a route towards reducing the burden of reporting on SMEs and finding the necessary factors which drive corporate carbon intensity.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Rosales, Charlyn UNSPECIFIED
Uncontrolled Keywords:	Algorithmic Carbon Accounting; XGBoost Regressor; Multi-Output Random Forest; K-Means Clustering; SME sustainability
Subjects:	G Geography. Anthropology. Recreation > GE Environmental Sciences Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Artificial Intelligence for Business
Depositing User:	Ciara O'Brien
Date Deposited:	02 Sep 2026 11:40
Last Modified:	02 Sep 2026 11:50
URI:	https://norma.ncirl.ie/id/eprint/9775

Actions (login required)

View Item