Classification Model Analysis of ICU Mortality Level using Random Forest and Neural Network

Abstract

Based on the results of previous studies, research on machine learning for predicting ICU patients is crucial as it can aid doctors in identifying high-risk individuals. A high accuracy in machine learning models is necessary for assisting doctors in making informed decisions. In this study, machine learning models were developed using two models, namely Random Forest and Artificial Neural Network (ANN), to predict patient mortality in the ICU. Patient data was obtained from The Global Open Source Severity of Illness Score (GOSSIS) and underwent preprocessing to address issues of missing values and imbalanced data. The data was then divided into training, validation, and testing sets for model training and evaluation. The results of the study indicate that the Random Forest model performs better with an accuracy of 93% on the testing data compared to the ANN which only achieved an accuracy of 86% on the testing data. Consequently, the Random Forest model can be utilized as a solution for predicting patient mortality in the ICU.

Keywords

• icu patient data analysis
• gossis (the global open source severity of illness score) dataset
• icu patient mortality prediction
• random forest
• neural network