An Ai-Based Intelligent Approach for Credit Risk Assessment

Abstract

Credit risk poses a substantial problem to the banking and financial industries, especially when borrowers fail to satisfy their repayment commitments. Conventional approaches have various challenges in effectively anticipating credit risk evaluations, including the incidence of fraudulent activity. Therefore, to avoid these problems, a new approach called the Pigeon U Net Prediction System (PUNPS) has been developed for credit risk prediction and classification. The credit card transaction dataset was collected using the Kaggle platform. The dataset was then preprocessed to remove duplicate items. The feature selection approach was used to keep only relevant variables. Credit risk prediction was successfully carried out using the fitness function of pigeon optimization. Furthermore, the classified credit risk forecasts were processed with the U-Net framework. Finally, the performance of the model was evaluated, and the findings were compared with standard approaches. This method offers significant advantages compared to conventional models, demonstrating improved performance in predicting credit risk through improved accuracy. The performance of this model is evaluated through various risk assessment metrics such as F1 score, Precision, recall, Accuracy, and error rate. It shows an impressive accuracy of 99.8%, accompanied by a precision score of 99.9% and a recall score of 99.7%. An F1 score of 99.6% confirms its effective balance between Precision and recall, establishing it as a reliable and accurate tool for credit risk assessment. In addition, it maintains a minimal error rate of only 0.2%.