Classification of Hypertension Severity Levels Using the Random Forest Method

Abstract

The rising prevalence of hypertension in Indonesia specifically in Samarinda City, which recorded 215,206 cases in 2023 poses challenges for the field of Informatics Engineering regarding the efficiency and consistency of manual disease-level classification in high-patient-volume healthcare facilities. The limitations of conventional approaches create a need for a technology-based system that can classify hypertension severity quickly, accurately, and objectively. This study aims to develop a hypertension severity classification model using the Random Forest algorithm optimized via GridSearchCV and to identify dominant risk factors through Feature Importance analysis. The dataset comprises 5,373 adult patient medical records from the Remaja Community Health Center (Puskesmas) in Samarinda, covering January to December 2024; attributes include gender, age, smoking status, height, weight, systolic and diastolic blood pressure, and a BMI feature derived through feature engineering. Data imbalance was addressed using the SMOTE technique on the training data. Results indicate that the model achieved the highest accuracy of 99.75% with a 70:30 data split, while an 80:20 ratio yielded 99.62% accuracy; both figures exceed the minimum target of 90%. Evaluation using accuracy, precision, recall, and F1-score metrics demonstrates the model's superiority over manual classification. Feature Importance analysis identifies diastolic blood pressure (46.31%) and systolic blood pressure (45.04%) as the most dominant factors, followed by age, BMI, and other attributes. The resulting model can be implemented as a clinical decision support system to assist medical personnel in primary healthcare settings.