Funnel Random Forest: Inliers-Focused Ensemble Learning for Improved Prognostics of Heart Failure

This study introduces a novel Funnel Random Forest (FRF) methodology to improve the accuracy and reliability of clinical prognostics, particularly for patients with heart failure. FRF identifies outliers using information theory-based metrics during training, reducing outliers by 11.04% compared to traditional methods. The model was trained on a dataset of 299 clinical records, achieving an 82.56% accuracy, a 1.67% improvement over standard Random Forest models. FRF outperformed existing models, with a 9.67% increase in accuracy and a 25.6% improvement in reliability, measured by the Precision-Recall Area Under the Curve (PR-AUC). These results demonstrate the model's potential to enhance clinical decision support, leading to better patient outcomes through early and precise prognostic predictions.