Penerapan Algoritma K-Means Untuk Klasterisasi Pasien Berdasarkan Data Penyakit

Abstract

Sibuhuan Community Health Center (Puskesmas Sibuhuan) experiences a continuous increase in patient visit data every month. However, the utilization of these data is still limited to administrative reporting, resulting in the underutilization of information regarding patient group characteristics. This study aims to implement the K-Means algorithm for patient clustering and to identify patient characteristic patterns based on age, gender, and disease variables. This research employed a quantitative method with a data mining approach. The data used were secondary data consisting of 1,000 patient visit records from Puskesmas Sibuhuan during the January 2026 period. Following data cleaning, 104 records containing missing values were removed, resulting in 896 records used for the clustering process. The research stages included data preprocessing, Min-Max normalization, determining the optimal number of clusters using the Elbow Method, clustering using the K-Means algorithm, and evaluating the clustering results using the Silhouette Coefficient. The results showed that the optimal number of clusters was three, Cluster 0 consisted of 423 patients, predominantly diagnosed with diseases of the respiratory, digestive, and circulatory systems. Cluster 1 comprised 320 patients, with the most prevalent diagnoses involving symptoms, signs, and abnormal clinical findings, as well as pregnancy, childbirth, and the puerperium. Cluster 2 consisted of 153 patients, predominantly diagnosed with infectious and parasitic diseases, endocrine, nutritional and metabolic diseases, and mental and behavioral disorders. The evaluation produced a Silhouette Coefficient value of 0.6384, indicating that the clustering results were of good quality. The findings of this study are expected to provide supporting information for Puskesmas Sibuhuan in identifying patient group characteristics and to assist in planning more targeted healthcare services.