Leveraging Large Language Models for Real-time Financial Crisis Identification

Details

Supervisors

Monti, Francesca

Faculty

Faculté des sciences économiques, sociales, politiques et de communication

Degree label

Master [120] en sciences économiques, orientation économétrie, à finalité approfondie

Abstract

This thesis examines the use of large language models (LLMs) for real-time financial crisis identification using textual data from OECD reports. The study explores the possibility of applying state-of-the-art models to textual data to solve some of the limitations which are attributed to traditional numerical models. Meta’s LLAMA-2 model was prompted to analyse the economic reports and decide whether a particular report described a period of a financial crisis. The model achieved an accuracy rate of over 91% and a ROC-AUC score of 0.836. Results demonstrate that LLMs can effectively identify financial crises, offering a valuable complement to numerical models. This research enhances the literature on NLP applications in economics and proposes a novel component, which could be added to early warning systems employed by central banks and policymakers.

Citations

• APA

Koveshnikov, V. (2024). Leveraging Large Language Models for Real-time Financial Crisis Identification. https://dial-mem.test.bib.ucl.ac.be/handle/2078.2/41798