CEU Electronic Theses and Dissertations, 2025
| Author | Shaken Abylaikhan |
|---|---|
| Title | Modelling Probability of Default Using Data on Hungarian Companies |
| Summary | This capstone project develops a machine learning framework to predict financial distress among Hungarian companies using firm-level accounting data from 2014 to 2023. Unlike traditional credit scoring models that rely on fixed formulas and rare default events, the project introduces a custom distress label based on four key financial ratios: profit to assets, EBITDA margin, revenue growth, and equity to assets. Firms in the bottom 30% of a standardized composite score are labeled as distressed. The model training and evaluation follow a strict out-of-sample design, with the training period covering 2014–2019 and the testing period set to 2022–2023. The study applies logistic regression, random forest, XGBoost, and a soft voting ensemble to predict distress, using non-leaky features such as liquidity, liabilities, and firm age. Evaluation metrics include ROC AUC, precision, recall, and F1-score. The final model, deployed in an interactive dashboard, enables financial analysts to assess distress probabilities in real time. This work demonstrates how modern data science methods can improve credit risk assessment, especially for small and medium-sized enterprises in Central and Eastern Europe. |
| Supervisor | Schindele Ibolya |
| Department | Economics MSc |
| Full text | https://www.etd.ceu.edu/2025/shaken_abylaikhan.pdf |
Visit the CEU Library.
© 2007-2025, Central European University