Abstract :

Predicting ICU admission using time series analysis plays a crucial role in improving early clinical decision making, optimizing resource allocation, and reducing mortality in critical care environments. This work focuses on building a predictive framework capable of analyzing physiological signals, electronic health records, and continuous monitoring parameters to identify patients at risk of ICU transfer. Time-dependent features such as vital signs, laboratory trends, and real-time patient variability are extracted and learned using advanced machine learning and deep learning models. The proposed system processes streaming healthcare data and predicts ICU admission probability with high reliability. Performance evaluation is carried out using benchmark critical care datasets. The results demonstrate improved sensitivity, accuracy, and predictive robustness. This research establishes a strong foundation for proactive healthcare support systems.