DEVELOPMENT OF PREDICTION MODEL FOR DIABETES USING BAYESIAN NETWORK

• Main Author: Marturia Sihombing, Paul
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77926
• Institution: Institut Teknologi Bandung

This research adopts a predictive approach to address the challenges of diabetes, with the primary aim of forecasting the probability of the occurrence of this disease. Indonesia faces a serious issue regarding diabetes, compounded by the difficulty in early prediction, exacerbating the situation. This disease has become a worrisome public health burden, underscoring the need for a more effective predictive solution. The study focuses on risk factors contributing to diabetes. Data for this analysis was obtained from the Centers for Disease Control and Prevention (CDC), encompassing comprehensive information on individuals' health characteristics and medical histories. The development of the prediction model is executed by adopting Bayesian Network, capable of integrating the complexity of risk factors into a structured predictive model. Experiments were carried out through the implementation of five predefined Bayesian Network scenarios. Each scenario reflects a unique combination of risk factors with the potential to influence the development of diabetes. The best experimental results indicate that this approach achieved an accuracy rate of 73% in predicting the probability of diabetes occurrence. The most influential dominant factors contributing to diabetes include age, body mass index, blood pressure, cholesterol levels, a history of heart disease, a history of stroke, and mobility difficulties. These findings offer crucial insights into the potential application of Bayesian Network in diabetes prediction based on relevant risk factors. This research not only highlights the serious challenges faced by Indonesia concerning diabetes but also provides impetus for addressing them with smarter and more accurate predictive solutions. Thus, this study takes a significant step in the prevention and management of diabetes in Indonesia.