HOSPITAL MANAGEMENT SYSTEM USING MACHINE LEARNING

Abstract

Hospital resource management and patient care depend on accurate ICU LOS prediction. A conventional machine learning-based LOS prediction framework is supplemented with the CatBoost technique, which effectively manages categorical EHR data and prevents overfitting. A dataset of hospital stays is used to train an extended model, and its accuracy, precision, recall, F1-score, and AUC are evaluated. In terms of prediction accuracy, CatBoost outperforms conventional classifiers. Explainable Artificial Intelligence (XAI) techniques such as SHAP are used to identify and assess clinical aspects that affect the length of an intensive care unit stay in order to improve model transparency. Through a web interface, Flask deployment provides real-time, intuitive ICU stay prediction. The proposed extension shows how CatBoost and explainable machine learning can accurately, understandably, and practically estimate the length of stay in a critical care unit in modern healthcare.