|
Predicting financial crises in Islamic banking
using machine learning and unbalanced data
|
Saeid MohammadBeigi *    |
| in Islamic Economic Philosophy, Imam Khomeini Educational and Research Institute |
|
|
Abstract: (19 Views) |
| This research was conducted with the aim of designing a financial distress prediction framework for Islamic banks, as existing models primarily focus on conventional banks and do not adequately address the specific characteristics of Islamic banking, such as Shariah requirements. The research method involved using financial, macroeconomic, and Shariah governance data from 450 bank-year observations (including both Islamic and conventional banks) from 2015 to 2023. The dependent variable, financial distress status, was defined based on the criteria of Laeven and Valencia (2018), adapted for Islamic banking. The main innovation of the research is the introduction and quantitative measurement of the variable "Fiqh Distance from AAOIFI Standards" as a predictor of distress. To address data imbalance (only 5.3% of observations pertained to distress), the SMOTE technique was used. Various machine learning algorithms, including Logistic Regression, Decision Tree, SVM, and XGBoost, were evaluated. Findings showed that the XGBoost algorithm, with a sensitivity of 0.83 and an AUC-ROC score of 0.93, performed best in identifying distressed banks. The "Fiqh Distance" variable was identified as the second most important variable after Return on Assets (ROA). Analyses confirmed that an increase in this distance (decreased compliance with Shariah standards) significantly raises the probability of distress occurring. Furthermore, the developed model for Islamic banks demonstrated higher diagnostic accuracy (sensitivity 0.87) compared to conventional banks (0.76). By filling a gap in the literature, this research provides an operational framework for bank supervisors (such as central banks and IFSB) to identify vulnerable Islamic banks earlier. The results indicate that adherence to Shariah governance and transparency is not only a religious obligation but also an effective risk-mitigation business strategy. |
|
| Keywords: Islamic banking, financial crisis prediction, machine learning, XGBoost, imbalanced data. |
|
|
Full-Text [PDF 1639 kb]
(9 Downloads)
|
|
Type of Study: Research |
Subject:
Special
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
