Predicting credit defaults is crucial for financial institutions to assess risk and make informed lending decisions. One of the most recent strategies banks and financial institutions have been testing to minimize losses that arise from credit default is the deployment of Early Warning Systems (EWS). By nature, this technique was primarily proposed and explored for commercial customers. However, this study proposes a comprehensive data-driven approach to model Early Warning Systems (EWS) for retail customers in the financial industry while using different Machine Learning (ML) models. We use Logistic Regression (LR), Gradient Boosting (GB), and Random Forest (RF) to classify customers' status, indicating the need to include potential default in a “watch list”. Additionally, we implement a fourth model (i.e., meta-model), whose predictions are based on the output of the other algorithms used (LR, GB, RF). Results indicate that the meta-model achieves higher accuracy than GB or any other individual model tested. From the management perspective, the findings indicate that a higher threshold for warning signals results in alerts closer to the overdue date, indicating increased sensitivity to emerging client deterioration. Conversely, lower thresholds focus more on the client's overall status. Furthermore, using the top ten features for training yields satisfactory overall results, but incorporating features beyond the top ten provides valuable supplementary information to be used in the decision-making process.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号