Development of financial distress prediction model for the watchlist classification of wholesale banking clients
Details
- Authors: Daniel Chen
- Title: Development of financial distress prediction model for the watchlist classification of wholesale banking clients
- Supervisior: Prof. Dr. Jörg Osterrieder
- Degree: Master of Science
- University: University of Twente
- Year: 2023
- Status: Working Paper
Summary
Development of financial distress prediction model for the watchlist classification of wholesale banking clients.
Abstract
This study presents a machine learning approach for developing a watchlist trigger to effectively monitor credit portfolios and detect clients in financial distress. The proposed financial distress prediction model incorporates historical internal and external early warning trigger data with internal client data from ING to forecast if clients will migrate to a watchlist or default status. The study evaluates several supervised learning algorithms and shows that the Random Forest model has the highest F1 score. The final watchlist trigger is evaluated based on three metrics - migration sensitivity, trigger precision, and time lag - to measure its predictive performance and timeliness. Conclusively, the introduced watchlist trigger outperforms the existing univariate early warning triggers, but more research is required to make conclusions about the timeliness. The research contributes to the literature by providing a case study at ING where a financial distress prediction model has been implemented at an individual institution level.