Heart Failure Patient Classification Using LogitBoost: Toward Robust Predictive Modeling
Thanyaphon Sommhay, Kittipat Noopromjaroen, Jirapipat Tantiwisan, Benjawan Khunprasert, Narumol Chumuang
Abstract
Heart failure (HF) is a leading cause of morbidity and mortality worldwide, with growing prevalence and substantial clinical challenges. Reliable risk prediction models are essential for improving early diagnosis and treatment strategies. This study develops a classification model for heart failure patients using the LogitBoost algorithm. The proposed framework includes data preprocessing, feature selection, and model development to enhance predictive accuracy and interpretability. Experiments were conducted on standardized HF datasets with training-testing splits of 50:50, 60:40, and 70:30. Performance was evaluated against multiple baseline classifiers using accuracy, precision, recall, F1-score, and AUC-ROC. Results demonstrate that LogitBoost achieved superior discriminative ability, with an accuracy of up to 85.89%, while maintaining computational efficiency on an Intel Core i5 processor. These findings highlight the robustness of LogitBoost in handling heterogeneous and imbalanced clinical data, underscoring its potential to support intelligent healthcare systems and improve patient outcomes.