Ensemble Learning Based Ventilator-Associated Pneumonia Prediction Within 24 Hours

Ventilator-Associated Pneumonia (VAP) is a significant issue in intensive care units, as it contributes to the rise in healthcare costs, morbidity, and mortality.Our research puts an effort into addressing the necessity of timely VAP prediction by creating and assessing delicate machine learning and deep learning models that can detect VAP one day prior to clinical diagnosis.A five-fold cross-validation scheme is employed to rigorously evaluate a number of models on a dataset extracted from the MIMIC-III database.The most effective model was a soft voting classifier that included k-nearest neighbors, support vector machine, gaussian naive bayes, decision tree, and logistic regression.It exhibits exceptional efficacy in distinguishing VAP from non-VAP instances, achieving an F1-score, precision, recall, and accuracy of 99.78%, 99.81%, 99.56%, and 99.78% respectively.The results prove its superiority and robustness in making an early prediction of VAP.