Devni Prima Sari, Yuni Wardana, Dina Agustina
Diabetes has emerged as a critical worldwide health issue, exhibiting a precipitous progression and the possibility of fatal complications if not effectively controlled. According to a report by the World Health Organization, diabetes impacted an alarming 422 million people globally in 2018, leading to an annual toll of 3.8 million fatalities. The International Diabetes Federation (IDF) has projected a concerning trajectory for Indonesia, which stands to experience a 47% increase in diabetes cases to 28.57 million by 2045. The IDF projects that the number of individuals with diabetes could reach a staggering 783.7 million on a global scale by the same year. This study utilizes Bayesian Network analysis to investigate complex associations between disease variables and factors that contribute to symptoms of diabetes on a publicly accessible diabetes dataset. The accuracy results of using the Bayesian Network method were 94%. 18% of the 390 entries in the dataset were associated with individuals who had already received a diagnosis of diabetes; the remaining 82% did not manifest the condition. Diabetes (X8), Cholesterol (X1), Gender (X2), Age (X3), BMI (X4), Glucose (X5), HDL cholesterol (X6), and Waist-to-Hip Ratio (X7) are the variables comprising this dataset. The utilization of the maximum spanning tree method to construct the Bayesian Network structure from complete weighted graphs is a significant contribution of this paper. This novel methodology improves our comprehension of the fundamental connections between the variables, ultimately enabling more precise prognostication of diabetes risk and proactive implementation of interventions. The findings highlight the potential of Bayesian Network analysis as a valuable instrument for predicting the risk of diabetes in its early stages. This presents a crucial opportunity to implement opportune interventions and manage the disease effectively, ultimately alleviating the burden associated with this potentially fatal condition. © 2026 Author(s).
Department of Mathematics, Universitas Negeri Padang, Padang, Indonesia
