Enhancing Bank Financial Performance Assessment: A Literature Review of Deep Learning Applications Using the Kitchenham Method

The assessment of bank financial performance is crucial for ensuring the stability of the banking sector. With advancements in technology, especially deep learning (DL), there is increasing potential to improve the accuracy of risk prediction and financial performance evaluation in banks. However, challenges related to data imbalance and model complexity require more efficient approaches. This study aims to examine the application of DL in assessing bank financial performance, with a focus on credit risk, fraud detection, and bankruptcy prediction. A Systematic Literature Review (SLR) was conducted using the Kitchenham approach, analyzing 697 relevant articles to address nine research questions regarding the implementation of DL in the banking sector. This study contributes by providing insights into effective DL models that enhance financial performance and risk prediction in banks, while also offering recommendations for the development of more transparent models. The results indicate that models such as Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN) perform well in handling large financial data. Additionally, hybrid models that combine DL with traditional models demonstrate higher accuracy in bankruptcy prediction and fraud detection.