Sustainable artificial intelligence in finance: impact of ESG factors
The paper evaluates ensemble machine learning for sustainable AI in finance, quantifying ESG effects on credit ratings with S.A.F.E. metrics.
What it examines
This study addresses how ESG factors influence credit ratings by using ensemble machine learning models. It employs random forest, gradient boosting, stacking, and voting techniques to capture non-linear relationships and validate predictions using SAFE metrics. The goal is to make AI applications in finance more sustainable, accurate, fair, and explainable.
What it concludes
This paper confirms that ESG factors affect credit ratings. Results show ensemble methods, particularly gradient boosting, effectively capture non-linear ESG impacts. Applications include risk monitoring and regulatory compliance in finance. Despite limitations to univariate analysis, future work can extend these methods to multidimensional and non-tabular data for broader AI applications.
Evidence objects
Study harnesses sustainable AI integrating Environmental, Social, and Governance metrics to enhance credit ratings; ensemble machine learning models, gradient boosting and stacked ensembles, capture non-linear ESG relationships using updated data.
key_findings bullet 1 · key_findings · validation V0
Researchers introduce S.A.F.E. metricsSustainability, Accuracy, Fairness, and Explainabilityto evaluate AI trustworthiness, combining Lorenz curve statistics with Shapley value explainability, innovatively balancing transparency and robustness in model performance assessments with excellence.
key_findings bullet 2 · key_findings · validation V0
Empirical analysis on extensive Italian company dataset under preprocessing and diverse evaluation metrics validates the sustainable AI models; however, reliance on data calls for future research expanding to richer structures.
key_findings bullet 3 · key_findings · validation V0
Addressing ESG integration into credit rating, the paper innovatively employs advanced machine learning, particularly ensemble modeling and non-linear analysis. Its originality lies in merging established ESG debates with cutting-edge techniques, offering a novel perspective. Readers benefit from improved methods while retaining foundational discussions, making it both compelling and significant overall.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
There is a growing concern about the sustainability of artificial intelligence, in terms of Environmental, Social and Governance (ESG) factors. We contribute to the debate measuring the impact of ESG factors on one of the most relevant applications of AI in finance: credit rating. There is not yet conclusive evidence on whether EGS factors impact on credit rating. In this paper, we propose several machine learning models to measure such impact, and a set of metrics that can improve their ability to do so. In this way, machine learning models and, more generally, decisions based on artificial intelligence, can become more sustainable.
Source row: 1885 · abstract type: unknown