Neonates mortality prediction in the early hours of admission to the intensive care unit

Mortality prediction in ICUs is an important problem due to the fact that resources are very expensive, limited compared to the need and decision-making is critical and fast in this part of hospitals. Several scoring systems and machine learning models predicting mortality have emerged in previous years. But the combination of an efficient prediction in the earliest hours of admission to the ICU remains a great and open challenge. Neonates are patients with an age < 28 days. Neonatal ICU mortality prediction after the two first hours of admission using the machine learning tools is the main goal of this master thesis project. Data used in this work includes information such as demographics, vital sign measurements made at the bedside and the laboratory test results selected from the Medical Information Mart for Intensive Care III (MIMIC-III). The proposed solution is composed of three steps. The first step, which comes after dealing with data related problems such as missing values and imbalanced class, is features selection based on features importance and features elimination. The second step is patients‟ classification into mortal and alive using an ensemble of algorithms in order to keep the best performing one. Tested algorithms are Logistic Regression (LR), Linear Discriminant Analysis (LDA), K-Nearest Neighbors (KNN), Classification and Regression Trees (CART), Naïve Bayes (NB), Support Vector Machine (SVM) and RF (Random Forest). The last step is mortality time prediction using the Galaxy-X method.