A Proposed Credit Rating Model for Rating Syrian Private Banks According to Creditworthiness Level Using Fuzzy Logic Enhanced by Genetic and Particle Swarm Optimization Algorithms
Keywords:
Syrian Private Banks, Creditworthiness, Credit Rating, Financial Stability, Fuzzy Logic, Genetic Algorithm, Particle Swarm Optimization Algorithm.Abstract
This research aims to develop a data-driven credit rating model for rating Syrian private banks according to their level of creditworthiness, using fuzzy logic enhanced by the Genetic and Particle Swarm Optimization Algorithms. The research relied, in building the model, on the financial stability indicator (Z-score) as a reference benchmark for assessing creditworthiness in the absence of official credit ratings for banks, while the model output represents a composite index reflecting banks’ creditworthiness, which is subsequently transformed into rating grades.
The model inputs were determined based on a set of bank-specific, sectoral, and macroeconomic variables with the highest ability to explain financial stability, which were extracted using the Genetic Algorithm and the Random Forest. Subsequently, a hierarchical fuzzy network was designed based on the Takagi–Sugeno framework, by transforming the inputs into intermediate nodes reflecting the internal and external environment of the bank, such that these nodes function as an aggregation layer linking the primary variables to the model output.
The empirical results demonstrated a high level of accuracy of the proposed model in representing the behavior of the reference financial stability indicator, with a coefficient of determination (R²) of 83.57%, while the root mean square error (RMSE) reached approximately 5.5%, which is a low level indicating the model’s ability to reduce the gap between the values generated by the model and the reference values. Moreover, the results revealed a strong and statistically significant positive rank correlation between the credit ratings derived from the model and the reference ratings, as measured by the tie-corrected Spearman’s rank correlation coefficient, which confirms the ability of the model to preserve the relative ordering of credit risks.
