In Hospital Mortality Risk Prediction for HF Patients Using SMOTE and Various Machine Learning Algorithms

Accurate prediction of mortality in ICU patients is crucial for timely intervention and appropriate medical services. Although various severity scores and machine learning models have been developed, accurate prediction of mortality remains challenging. This study aimed to address this challenge by validating the performance of commonly used classification models, namely Linear Discriminant Analysis, K-Nearest Neighbor, XGBoost, Decision Tree, and Random Forest, along with the Synthetic Minority Oversampling Technique (SMOTE) for predicting ICU mortality. The study utilized a comprehensive dataset consisting of 1177 cases from the MIMIC-III database, which encompasses a wide range of patient characteristics and clinical variables technique for predicting ICU mortality. The study was conducted on 1177 cases from the MIMIC-III dataset. Results showed that the proposed models outperformed state-of-the-art approaches with 100% accuracy, F1 score, precision, recall, and AUC-ROC. This study demonstrates the effectiveness of these models for predicting ICU mortality.