Enhancing credit risk evaluation with deep neural networks and ensemble learning techniques

Credit risk evaluation is an important topic for financial institutions. In this paper, we introduce a Deep Neural Network approach (DNN) boosted by two ensemble learning models: XGBoost (XGB) and Random Forest (RF). The experiment was conducted on a Kaggle competition dataset, consisting of 59 features (After data preparation) with vast data types. The DNN-based model demonstrated improved performance compared to XGB and RF, particularly in key-metrics. The study confirms the superiority of DNN in tasks that require high accuracy and efficiency, making it a viable option for integrating advanced neural architectures in real-world scenarios involving credit risk assessment.