Heart Failure Patients In-Hospital Mortality Risk Prediction Models Based on Regularization Techniques-Comprehensive Analysis

Any type of cardiovascular disorder tends towards Heart Failure (HF) at an ultimatum. The major shortfall of critical care units in a hospital can be handled by predicting the mortality risk of HF patients in intensive care units (ICUs) to provide personalized treatment plans and care for life prevention. The eleven Machine Learning (ML) classifiers are deployed for the initial mortality prediction of overall features. The two regularization techniques for the automatic feature selection of regression models like Lasso and Ridge with penalty L1 and L2 are used to reduce the features to find the importance of each feature. The observable accuracy achieved by the Light Gradient Boosting Machine (LGBM) classifier, among others, is 95.59% for L1 and 93.14% for L2. The feature selection decreases the model training time and increases the efficiency along with low resource costs. This provides helping the model fit properly, reducing complexity and overfitting risk.