Banks play a pivotal role in the global economy by channelling funds from savers to borrowers, thus facilitating business activities and promoting economic growth. Consequently, systemic banking crises can severely disrupt financial markets and hinder broader economic stability. This study aims to enhance the early detection of bank distress within the Indonesian banking sector, covering the period from 2014 to 2023. Data were gathered from publicly listed banks on the Indonesia Stock Exchange (IDX). Four models were employed to develop predictive frameworks: the Modified Altman model, the Ohlson model, the Zmijewski model, and the XGBoost model. The performance of each model was assessed using Overall Predictive Accuracy (OPA), Type I errors (misclassifying healthy banks as distressed), and Type II errors (failing to detect distressed banks). Among these, the XGBoost model emerged as the most effective, achieving the highest OPA and completely eliminating Type II errors. In contrast, the traditional accounting-based models displayed lower accuracy rates and more pronounced Type II errors, indicating a limited capacity to capture the complexities of modern banking data. These findings underscore the growing potential of machine learning algorithms, such as XGBoost, to substantially improve financial distress prediction and foster a more resilient banking system. By providing banks, regulators, and policymakers with robust early warning signals, the model proposed in this study helps mitigate the risks associated with bank failures, thereby enhancing financial stability and supporting sustained economic resilience in Indonesia and beyond. Additionally, future research may refine these models by incorporating macroeconomic indicators and recalibrating traditional approaches to reflect the evolving nature of the banking landscape.