PREDICTION OF HYPERTENSION COMORBIDITIES USING THE RANDOM FOREST ALGORITHM (CASE STUDY AT PUSKESMAS KASIHAN 2)

Abstract

Hypertension is a chronic non-communicable disease often accompanied by comorbidities, which can increase complication risks and reduce patients' quality of life. Currently, the identification of comorbidities in hypertension patients is frequently conducted manually, making it time-consuming and highly dependent on healthcare workers' thoroughness. This study aims to develop a predictive model for hypertension comorbidities using a machine learning-based Random Forest algorithm, designed as an early screening tool for the general population. The research method follows the Cross-Industry Standard Process for Data Mining (CRISP-DM) framework, encompassing business understanding, data understanding, data preparation, modeling, evaluation, and deployment. Clinical data were collected from medical records, focusing on predicting eight primary comorbidities using a multi-label classification approach. Data preprocessing involved data cleaning, transformation, splitting into training and testing sets, and handling class imbalances. The Random Forest model was trained and evaluated using subset accuracy and hamming loss metrics. The results demonstrate that the Random Forest algorithm successfully predicts hypertension comorbidities with a subset accuracy of 0.3361 and a hamming loss of 0.1502, indicating robust performance for multi-label prediction. The model was successfully deployed into a Streamlit-based web application, enabling healthcare professionals to obtain direct prediction results. This system is expected to assist in the early screening and monitoring of hypertension patients.