PERBANDINGAN KINERJA ALGORITMA MACHINE LEARNING UNTUK KLASIFIKASI ISPA MENGGUNAKAN DATA KLINIS RUMAH SAKIT

Abstract

Acute Respiratory Infection (ARI) remains one of the leading causes of morbidity and mortality worldwide, particularly among children and elderly populations. The complexity of ARI clinical symptoms necessitates rapid and accurate diagnostic approaches to support healthcare services. This study compares the performance of five machine learning algorithms, Logistic Regression, Naïve Bayes, K-Nearest Neighbor, Random Forest, and Gradient Boosting Machine algorithms for ARI classification using clinical hospital data. The study employed a quantitative experimental approach, using 521 outpatient clinical records obtained from XYZ Hospital, Jakarta. The research process included data preprocessing, classification model development, model performance evaluation, and statistical analysis using the Kruskal-Wallis test followed by Dunn's Post Hoc Test with Bonferroni correction. Model performance was assessed using accuracy, precision, recall, and F1-score metrics, which were computed using macro averaging due to the imbalanced class distribution. Statistically significant differences were observed among the algorithms across all evaluation metrics (p < 0,001). Effect size analysis using epsilon squared (ε²) indicated large effects for accuracy (ε² = 0.828), precision (ε² = 0.719), recall (ε² = 0.434), and F1-score (ε² = 0.654). The post hoc analysis indicated that Random Forest and Gradient Boosting Machine showed comparable performance and consistently achieved competitive results across evaluation metrics. These findings suggest that ensemble learning methods are better suited to handling the complex clinical data associated with ARI and could help develop decision support systems for early ARI screening. Future studies should incorporate multicenter datasets, hyperparameter optimization, and explainable artificial intelligence techniques to improve model generalizability and interpretability.