The boiler station is a critical section in palm oil mills as it supplies steam for sterilization and energy generation. However, a case study revealed that the boiler station operated with an Overall Equipment Effectiveness (OEE) of only 27.03%, far below the 85% world-class standard. The main contributors to this inefficiency were reactive maintenance practices, delayed failure responses, and aging components with no early detection mechanisms, resulting in frequent breakdowns and prolonged downtimes. This research was conducted with three objectives: (1) to evaluate and analyze the root causes of low OEE performance, (2) to develop a predictive maintenance system using logistic regression to forecast component failures based on historical maintenance data, and (3) to integrate an automated alert system that notifies engineers of high-risk components before failures occur. Logistic regression was selected for its simplicity, interpretability, and suitability for environments with limited digital infrastructure, making it ideal for palm oil mills with structured Excel-based records. The predictive system was developed using Python and Streamlit, enabling engineers to upload maintenance logs, receive failure probability predictions, and visualize high-risk components through an interactive web interface. The model achieved a prediction accuracy of 77%, demonstrating its effectiveness in shifting maintenance practices from reactive to predictive, improving equipment reliability, and aligning with Total Productive Maintenance (TPM) and Industry 4.0 principles.