Intelligent Medical Decision Support System for Predicting Patients at Risk in Intensive Care Units


Patients’ lives can be rescued by a prediction made by an Intelligent Medical Decision Support System (IMDSS). Such a system can harness the information wealth of patient Electronic Medical Records and leverage up-to-date Machine Learning technology. The accuracy of prediction is one of the most critical characteristics of this intelligent system. Moreover, the technical issues of the medical data as the curse of dimensionality and imbalance are significant challenges. In this paper, we implement the main building block of an IMDSS which is the predictive model. In addition, a comprehensive study of different accuracy factors of this system is given. We tested different approaches and methods for these factors to reach an optimal setting for system development. A big real-world medical dataset is used to test the model for predicting the in-hospital risk of mortality from only the first 24 hours of stays in the Intensive Care Unit.

This document is currently not available here.