Comparative Analysis of Mortality Prediction Models at the University Hospital Center of Oran, Algeria

Abstract

Predicting mortality is an important field of study that aids in making wise healthcare decisions and offers insightful information about population health.  Using demographic and hospital-service data from the University Hospital Center of Oran (CHUO), Algeria, this study employs machine learning (ML) models to forecast the ultimate causes of mortality.  Sex, city of residence, hospital services used, and the beginning, intermediate, and ultimate causes of death are among the factors included in the 12.604 records that make up the dataset.  To find trends and forecast the causes of death in eight distinct groups, six machine learning models—Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), Naive Bayes (NB), Multilayer Perceptron (MLP), and Extreme Gradient Boosting (XGBoost) were trained and assessed. XGBoost achieved an accuracy and specificity of 84.05%, with a precision of 42.73%, recall of 25.53%, and an F1 score of 28.33%, the model outperformed the other evaluated models, proving its ability to effectively capture intricate relationships in the data.  The study demonstrates how machine learning techniques can be used to examine a variety of variables and find significant patterns in mortality trends.  This work enhances predictive analytics in healthcare by utilizing local data and sophisticated algorithms, providing useful instruments for directing public health initiatives.  The results highlight how machine learning can improve healthcare outcomes and solve issues connected to mortality in Algeria.