Heart Disease Risk Classification and Feature Importance Based on Machine Learning Using Microdata SKI 2023

Abstract

Cardiovascular disease remained the leading cause of mortality in Indonesia, with death rates increasing by more than 25% and national health expenditure reaching Rp17.92 trillion. Its complex risk profile requires population-based predictive approaches. The 2023 Indonesian Health Survey (SKI) provided extensive data suitable for Machine Learning-based risk modelling. This study aimed to develop a classification model for heart-disease risk and identify dominant risk factors using the Random Forest algorithm. A case-control design was applied, with data divided into training and testing sets using an 80:20 ratio to ensure objective model evaluation. The analysis followed the Knowledge Discovery in Database (KDD) framework, including data selection, preprocessing, transformation, modelling, and evaluation. Random Forest was used for classification, while feature importance was assessed using Information Gain and Gain Ratio. Model performance was evaluated using accuracy, sensitivity, and specificity. Age, hypertension, and Body Mass Index (BMI) were identified as the most influential predictors. The model achieved an accuracy of 72.13%, sensitivity of 73.11%, and specificity of 71.26%, indicating stable classification performance on large population data. However, this study is limited by the use of secondary data and the absence of external validation. These findings highlight the potential of Machine Learning to support population-based risk stratification and inform targeted prevention strategies, contributing to evidence-based policy development and early screening programs in primary healthcare settings.