امیر خرمی

Objective : This paper introduces a groundbreaking optimization-simulation model, a novel approach that promises to revolutionize credit scoring and loan optimization for banks. Methods : The proposed approach follows a three-stage framework: data preparation, credit scoring, and optimization simulation. In the data preparation stage, corporate client data, including bank loan information and financial statements, has been collected and processed to define and calculate relevant features. The credit scoring stage involved meticulous feature selection using the correlation method, followed by the rigorous training and testing of five classification methods: logistic regression (LR), K-nearest neighbors (KNN), artificial neural network (ANN), adaptive boosting (AdaBoost), and random forest (RF). Model performance has been evaluated using accuracy, F1-score, and area under the curve (AUC) to identify the most effective classifier. In the optimization-simulation stage, the Memetic Algorithm (MA) has been utilized to optimize loan characteristics, including loan size, interest rate, and repayment period, while minimizing the rate of loan defaults. Additionally, this stage incorporated the pre-trained credit scoring model to estimate the impact of loan characteristics on default probabilities.  Results : A case study was conducted using data from 1,000 corporate clients of Bank Tejarat. The optimization-simulation approach has successfully reduced the loan default rate from 33% to below 5%, a significant achievement that underscores its potential to mitigate banks' credit risk. This shows the effectiveness of the proposed method in reducing credit risk for banks. Additionally, the AdaBoost technique achieved the best performance among the credit assessment models. Conclusion : The optimization-simulation approach combines determining the optimal loan specifications with the credit assessment process. This approach considers the impact of loan characteristics on the likelihood of customer default and utilizes this information to reduce banks' credit risk