Seleksi Fitur Berbasis Mutual Information untuk Optimalisasi Model Prediksi Tingkat Kematian Penderita Gagal Jantung Menggunakan Machine Learning

Abstract

Heart failure is one of the leading causes of death in the world that requires an accurate and efficient prediction system to support clinical decision making. This study aims to develop a prediction model for the risk of death in patients with heart failure by optimizing feature selection using the Mutual Information (MI) approach. The main problem raised is the high complexity of clinical data with many features that are not always relevant, which can reduce the accuracy and efficiency of predictive models. The method proposed in this study involves MI-based feature selection to identify the most informative features against the target variable (patient mortality), which are then used to train various machine learning algorithms such as Random Forest, Gradient Boosting, XGBoost, and Logistic Regression. The hyperparameter tuning process is performed to optimize the performance of each model. The test results show that the Random Forest model that has been tuned using five selected features managed to achieve an accuracy of 0.99 and F1-score of 0.99, outperforming other models in terms of balance between accuracy and generalization. The results show that Mutual Information is effective in simplifying model complexity without compromising prediction performance.