Pemetaan Pola Risiko Obesitas Berdasarkan Indeks Massa Tubuh dan Aktivitas Fisik Menggunakan Jaringan Kohonen (Self-Organizing Map)

Abstract

The increasing number of non-communicable diseases, such as diabetes mellitus and hypertension at the primary healthcare level, is closely linked to the rising trend of obesity. Thus far, the utilization of health data—including height, weight, and physical activity—at community health centers (Puskesmas) has been limited to individual record-keeping and has not been collectively analyzed to map high-risk populations. This study aims to implement the Kohonen Network or Self-Organizing Map (SOM) algorithm to objectively map obesity risk patterns through unsupervised learning without predefined class labels. The research dataset was obtained from the Pematang Pasir Community Health Center in Tanjungbalai, North Sumatra, comprising 200 individual medical records with attributes including Age, Body Mass Index (BMI), Frequency of Physical Activity (FAF), and Number of Meals Consumption (NCP). The dataset was normalized using the Z-Score method prior to being trained using the Python-based SOM algorithm. The results successfully grouped the community's obesity risk patterns into three final clusters: High-Risk Cluster, Medium-Risk Cluster, and Low-Risk Cluster. This visual mapping can be utilized by the health center authorities to design more targeted preventive health programs at the community level.