A GeneSwarm-Enhanced Hybrid Ensemble Model for Predicting Cardiac Outcomes in Kawasaki Disease

Kawasaki Disease (KD) is a paediatric vasculitis that may cause coronary artery lesions (CAL) in case it is not diagnosed and treated early. The problem of complicated clinical data and nonspecific markers has complicated proper forecasting of the long-term KD outcomes. This paper suggests a hybrid ensemble machine learning model to use XGBoost, AdaBoost, and Support Vector Machine (SVM) classifiers and soft voting in order to have robust and interpretable predictions. The GeneSwarm Feature Selector was used to select the features based on the Genetic Algorithms and Particle Swarm Optimization which can optimally identify the most informative and non-redundant clinical and laboratory features. The Babysitting algorithm was used to optimize hyperparameters dynamically and Interquartile Range (IQR) to eliminate outliers and Synthetic Minority Oversampling Technique (SMOTE) to deal with class imbalance were used during preprocessing. Assessed on 1,850 paediatric KD cases, the proposed model demonstrated a 97.10% accuracy, 96.50% precision, 96.80% recall, 96.60% F1-score, and a macro-AUC of 97.40% and is better than the current ML methods. Best predictive variables were Complications, Clinical Outcomes, Treatment Approach, Echocardiography, Laboratory Tests, Fever Duration, and Symptoms. These findings indicate that the framework is a reliable, interpretable, and clinically applicable instrument of early KD prognosis, which helps to plan personalized treatment and stratify risks. Through the incorporation of superior feature selection, ensemble learning, and a solid preprocessing, the research will add to an empirical and pragmatic method of enhancing cardiac care in pediatrics.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.