Publicação
Traditional Machine Learning vs. GPT-4: A Comparative Study on ESG Score Prediction
| datacite.subject.fos | Ciências Naturais::Ciências da Computação e da Informação | pt_PT |
| dc.contributor.advisor | Rio, José Américo Alves Sustelo | |
| dc.contributor.author | Ciraso, Chiara | |
| dc.date.accessioned | 2024-11-07T11:41:42Z | |
| dc.date.available | 2025-10-29T01:30:24Z | |
| dc.date.issued | 2024-10-29 | |
| dc.description | Dissertation presented as the partial requirement for obtaining a Master's degree in Statistics and Information Management, specialization in Information Analysis and Management | pt_PT |
| dc.description.abstract | This thesis examines the performance of traditional machine learning (ML) methods and large language models (LLMs), specifically GPT-4, within the context of predicting Environmental, Social and Governance (ESG) scores. Leveraging on a dataset based on information by leading companies in the realm of environmental indicators and ESG rating, respectively Urgentem and Moody’s, this study evaluates the models under three distinct scenarios: high-dimensional noisy data, lowerdimensional imbalanced data, and datasets with scarce information. Traditional ML methods, including Decision Trees (DT), K-Nearest Neighbors (KNN), and Support Vector Machines (SVM), are compared against GPT-4's capabilities in handling these conditions to explore strengths and weaknesses of both approaches across different aspects of the analytical chain, from data pre-processing to interpretation. Results indicate that traditional ML models, particularly ensembles of DT, KNN, and SVM, excel in handling high-dimensional and noisy. In contrast, GPT-4 demonstrates superior performance with simpler, well-defined use-cases. Both approaches reveal specific advantages and limitations, highlighting the importance of model selection based on dataset characteristics and specific application needs. The thesis underscores the importance of balancing automated approaches with expert control. The findings suggest a complementary application of traditional ML methods and LLMs, leveraging their respective strengths for more robust and comprehensive ESG scoring frameworks tailored to the practitioners’ need. | pt_PT |
| dc.identifier.tid | 203776224 | pt_PT |
| dc.identifier.uri | http://hdl.handle.net/10362/174754 | |
| dc.language.iso | eng | pt_PT |
| dc.subject | Machine Learning | pt_PT |
| dc.subject | Large Language Models | pt_PT |
| dc.subject | Generative AI | pt_PT |
| dc.subject | GPT-4 | pt_PT |
| dc.subject | ESG Scores | pt_PT |
| dc.subject | Sustainability | pt_PT |
| dc.subject | SDG 8 - Decent work and economic growth | pt_PT |
| dc.subject | SDG 9 - Industry, innovation and infrastructure | pt_PT |
| dc.subject | SDG 13 - Climate action | pt_PT |
| dc.subject | SDG 16 - Peace, justice and strong institutions | pt_PT |
| dc.title | Traditional Machine Learning vs. GPT-4: A Comparative Study on ESG Score Prediction | pt_PT |
| dc.type | master thesis | |
| dspace.entity.type | Publication | |
| rcaap.rights | openAccess | pt_PT |
| rcaap.type | masterThesis | pt_PT |
| thesis.degree.name | Mestrado em Estatística e Gestão de Informação, especialização em Análise e Gestão de Informação | pt_PT |