Sentiment-Driven Credit Risk Analysis: Hybrid Framework with Country-Level Fusion and Explainable AI

Acharekar, Ojas Shivadatta (2025) Sentiment-Driven Credit Risk Analysis: Hybrid Framework with Country-Level Fusion and Explainable AI. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (885kB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

The statistical probability that the borrower may not meet the financial obligations, alerting the lenders and financial institutions before making a decision, is known as credit risk. Traditional credit risk models that were developed mainly depended on structured data such as income, credit history, and employment status, but often neglected the dynamic behavioral signals that can act as an indicator in the early stages of financial instability. In this study, this gap is being addressed by introducing a hybrid framework that integrates the structured financial attributes with the sentiment data extracted from unstructured financial tweets. The Deep Learning models (LSTM and BiLSTM) were used with GloVe embeddings for sentiment classification, which achieved an accuracy of 92.28% suggesting validation on user link dataset. A novel aspect in this study is that the structured and unstructured data are integrated based on the country-level mapping, which ultimately enriches the individual credit profiles by adding the aggregate national sentiment scores. The merging of the two datasets was done by using the country column in both datasets. In addition to this, Explainable AI (LIME) was used to interpret the predictions of the model and find the correlation between sentiment and creditworthiness. As per the findings, the Stacking ensemble model achieved an accuracy of 83.66% having the best performance. Overall, considering the financial sentiments, an approach that comprises a scalable and explainable way of assessing individual credit profiles was introduced, which is suitable and more useful in real-time financial applications.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Ain, Qurrat Ul UNSPECIFIED
Uncontrolled Keywords:	Credit Risk; Sentiment Analysis; LSTM; Machine Learning; Financial Tweets
Subjects:	H Social Sciences > HG Finance 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 H Social Sciences > HG Finance > Credit. Debt. Loans. Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	30 Jun 2026 16:48
Last Modified:	30 Jun 2026 16:48
URI:	https://norma.ncirl.ie/id/eprint/9408

Actions (login required)

View Item