OPTIMASI ALGORITMA NAIVE BAYES UNTUK KLASIFIKASI PASIEN DEMAM BERDARAH DENGUE DI RSU SEBENING KASIH

OPTIMIZATION OF THE NAIVE BAYES ALGORITHM FOR CLASSIFICATION OF DENGUE HEMORRHAGIC FEVER PATIENTS AT SEBENING KASIH GENERAL HOSPITAL

Abstrak

Dengue Hemorrhagic Fever (DHF) remains a major public health issue in Indonesia, requiring rapid and accurate diagnosis to prevent severe complications. RSU Sebening Kasih, as a referral hospital in Pati Regency, faces challenges in classifying DHF patient care due to the presence of incomplete medical records (missing values). This study aims to optimize the Naive Bayes algorithm for DHF patient care classification by handling missing values using the Mean Imputation method. The dataset, obtained from the Hospital Management Information System (SIMRS) of RSU Sebening Kasih, consists of 635 patient records. The research process includes data preprocessing, handling missing values, data splitting into training and testing sets, and model evaluation using a Confusion Matrix based on accuracy, precision, recall, and F1-score metrics. After performing data cleaning, converting categorical attributes to numerical values, and imputing missing data, the Naive Bayes model was tested using an 80:20 stratified sampling ratio and evaluated with the aforementioned metrics. The results indicate excellent classification performance, achieving 98.43% accuracy in RapidMiner and 100% accuracy in Google Colab, with an average F1-score of 99.13% for inpatient cases and 91.67% for outpatient cases. These findings demonstrate that appropriate imputation methods can significantly enhance the performance of the Naive Bayes algorithm in classifying DHF patient care status. This research is expected to serve as a foundation for developing machine learning–based medical decision support systems at RSU Sebening Kasih and other hospitals to improve diagnostic effectiveness and healthcare service quality.