نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسندگان English
Objective: The primary objective of this study is to evaluate the performance of various deep learning algorithms, including DNN, CNN, and RNN, in accurately analyzing and predicting credit risk, as well as identifying complex patterns within banking data.
Method: This study utilizes a banking dataset comprising information on loans and customer accounts. The data were fully preprocessed, and subsequently, deep learning models including DNNs, CNNs, and RNNs were trained to predict credit risk. Training and optimization techniques such as optimization algorithms and regularization methods were applied to each model. The performance of the models was assessed using key metrics such as accuracy, sensitivity, specificity, and the F1 score.
Results: The results indicate that:
The RNN model achieved the best performance, with an accuracy of 88% and an F1 score of 88%, effectively capturing temporal patterns and periodic variations in the banking data.
- The DNN model obtained an accuracy of 85% and an F1 score of 84%, demonstrating strong capability in identifying complex patterns, although it was less effective in handling sequential data.
- The CNN model reached an accuracy of 82% and an F1 score of 81%, performing well in extracting local features, but its results were inferior to those of the DNN and RNN models.
Conclusions: The study demonstrates that RNNs, due to their superior ability to analyze temporal patterns, are the most effective model for predicting and mitigating credit risk in banking data. Although DNNs and CNNs also exhibit acceptable performance, they show certain limitations compared to RNNs in specific contexts. Therefore, selecting an appropriate model should be based on the characteristics of the data and the prediction requirements, in order to enhance credit risk management processes.
کلیدواژهها English
