Penerapan Algoritma Naive Bayes Dalam Memprediksi Kebutuhan Pelayanan Kesehatan Pasca Bencana Alam Di Sumatera Utara
Abstract
ABSTRACT
Abstract— Natural disasters can significantly increase the demand for health services while simultaneously disrupting the capacity of health facilities. This condition requires a simple predictive approach that can provide early information for post-disaster health response planning. This study evaluates the application of Gaussian Naive Bayes to classify post-disaster health service needs in North Sumatra using disaster-impact indicators and health-facility capacity. The dataset consists of 17 historical disaster records covering injured victims, fatalities, displaced residents, damaged health facilities, active hospitals, and active community health centers. The data were stratified into 11 training records and 6 testing records. Model performance was evaluated using accuracy, weighted precision, weighted recall, weighted F1-score, macro-averaged metrics, and a multiclass confusion matrix. The testing results yielded an accuracy of 83.33%, weighted precision of 70.83%, weighted recall of 83.33%, and weighted F1-score of 76.19%. However, class-level evaluation showed that the only High-need observation in the test set was misclassified as Medium. These findings indicate that Gaussian Naive Bayes has potential as a low-computational-cost component of a decision support system, but the evidence remains preliminary because of the limited dataset size. Future research should expand the dataset, improve the representation of each target class, and conduct validation using larger datasets before operational deployment.
Keywords: Health Services; Machine Learning; Naive Bayes; Natural Disasters; Prediction
Abstrak
Abstrak— Bencana alam dapat meningkatkan kebutuhan pelayanan kesehatan secara signifikan dan mengganggu kapasitas fasilitas kesehatan. Kondisi tersebut menuntut pendekatan prediktif yang sederhana dan dapat digunakan sebagai informasi awal dalam perencanaan respons kesehatan pascabencana. Penelitian ini bertujuan mengevaluasi penerapan Gaussian Naive Bayes untuk mengklasifikasikan tingkat kebutuhan pelayanan kesehatan pascabencana di Sumatera Utara berdasarkan indikator dampak bencana dan kapasitas fasilitas kesehatan. Dataset penelitian terdiri atas 17 catatan historis bencana yang mencakup jumlah korban luka, korban meninggal, penduduk mengungsi, fasilitas kesehatan rusak, rumah sakit aktif, dan puskesmas aktif. Data dibagi secara terstratifikasi menjadi 11 data latih dan 6 data uji. Evaluasi dilakukan menggunakan accuracy, weighted precision, weighted recall, weighted F1-score, serta metrik macro dan confusion matrix. Hasil pengujian menunjukkan accuracy sebesar 83,33%, weighted precision 70,83%, weighted recall 83,33%, dan weighted F1-score 76,19%. Namun, evaluasi per kelas menunjukkan bahwa satu-satunya observasi kelas Tinggi pada data uji salah diklasifikasikan sebagai Sedang. Temuan ini menunjukkan bahwa Gaussian Naive Bayes berpotensi digunakan sebagai komponen awal sistem pendukung keputusan dengan kebutuhan komputasi rendah, tetapi hasilnya masih bersifat awal karena ukuran dataset terbatas. Penelitian lanjutan perlu memperluas dataset, memperkuat representasi setiap kelas, dan melakukan validasi menggunakan data yang lebih besar sebelum diterapkan pada lingkungan operasional.
Kata Kunci: Pelayanan Kesehatan; Machine Learning; Naive Bayes; Bencana Alam; Prediksi
Downloads
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Abel Johannes Saragih, Rohitta Girsang, Guddes Theresia Naibaho, Gita Valentina Br. Tarigan, Jijon Raphita Sagala

This work is licensed under a Creative Commons Attribution 4.0 International License.





