Bank’s Credit Portfolio Optimization Using Actuarial Approach and Artificial Neural Networks

Objective
Allocating funds to various economic sectors and extending credit are among the key activities of banks. While following monetary and fiscal policies set by governments and central banks, banks strive to allocate these resources to profitable and suitable sectors. Credit risk reduction and control play a vital role in enhancing the lending process and, in turn, bank performance. Banks consistently pursue the dual objectives of minimizing risk and maximizing profit. Insufficient attention to credit yield and risk has led to the concentration of loans in specific economic sectors, creating significant challenges for banks. Considering the need to establish an optimal credit allocation and adopt effective policies, this study aims to develop an optimal model for credit allocation across economic sectors. By integrating actuarial methods and artificial neural networks (ANN) and considering banking policy constraints, the study seeks to design a credit portfolio that minimizes credit risk.
 
Methods
The actuarial approach, which includes calculating loan default probabilities and conducting precise financial risk assessments, is a widely used tool in bank risk management. However, these methods often fail to fully capture the complexities inherent in credit interactions. To address this limitation, this study incorporates artificial neural networks (ANN), which provide enhanced predictive accuracy and adaptability to nonlinear data. This research begins by analyzing the credit risk of the bank's loan portfolio using an actuarial approach and subsequently applies a perceptron neural network model to determine the optimal credit portfolio composition considering the bank's lending constraints. The sample data comprises loans extended by the bank to 280 major clients across four sectors—industrial, trade and services, agriculture, and construction—in 2013.
 
Results
The results indicate that the optimized portfolio, with a greater focus on the agricultural sector, can offer improved risk-adjusted returns compared to the bank’s current portfolio, which predominantly emphasizes the industrial sector. In the optimized portfolio composition, the agricultural sector receives the largest allocation, followed by the trade and services, construction, and industrial sectors. In contrast, in the bank’s existing portfolio, the highest allocation is to the industrial sector, followed sequentially by trade and services, agriculture, and construction. A review of the banking system's loan portfolio in 2013 confirmed the empirical validity of the model’s results.
 
Conclusion
Based on the findings and validation of the research hypotheses, it can be concluded that utilizing an actuarial model to determine credit risk, followed by optimization through artificial neural networks, enhances the bank’s credit portfolio optimization process. This approach enables banks to improve portfolio structure, which helps in mitigating potential risks and achieving more stable returns.