Scholarlink

Data play a crucial role in decision making for modern businesses, and the reliability of these decisions highly depends on the data quality. This problem is particularly relevant when we deal with financial data used for risk assessment and reporting. Basel regulations mandate that banks hold a certain amount of capital based on the level of risk in their portfolios expressed in terms of risk-weighted assets (RWAs). The quality of a bank’s risk management is directly impacted by the quality of the asset data used to calculate RWAs. In this study, we present a data quality (DQ) framework and show how machine learning paired with eXplainable Artificial Intelligence (XAI) techniques can be used to perform automatic DQ monitoring using a human-in-the-loop approach in credit risk. By obtaining expert feedback about model output enriched with XAI explanations, we clearly demonstrate the power of ML in terms of enhancing credit risk data quality and the advantages of using XAI to assist experts in analysing model outputs.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Boosting Credit Risk Data Quality Using Machine Learning and eXplainable AI Techniques

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Elena Tiukhova,
  • Adriano Salcuni,
  • Can Oguz,
  • Marcella Niglio,
  • Giuseppe Storti,
  • Fabio Forte,
  • Bart Baesens,
  • Monique Snoeck

摘要

Data play a crucial role in decision making for modern businesses, and the reliability of these decisions highly depends on the data quality. This problem is particularly relevant when we deal with financial data used for risk assessment and reporting. Basel regulations mandate that banks hold a certain amount of capital based on the level of risk in their portfolios expressed in terms of risk-weighted assets (RWAs). The quality of a bank’s risk management is directly impacted by the quality of the asset data used to calculate RWAs. In this study, we present a data quality (DQ) framework and show how machine learning paired with eXplainable Artificial Intelligence (XAI) techniques can be used to perform automatic DQ monitoring using a human-in-the-loop approach in credit risk. By obtaining expert feedback about model output enriched with XAI explanations, we clearly demonstrate the power of ML in terms of enhancing credit risk data quality and the advantages of using XAI to assist experts in analysing model outputs.