Peningkatan Prediksi Kelainan Tekanan Darah dengan Logistic Regression dan Random Forest: Pendekatan Sequence Machine Learning

Abstract

Early detection of blood pressure abnormalities plays a critical role in preventing and managing cardiovascular diseases, which remain the leading cause of death globally. This study proposes a sequence machine learning approach that combines Random Forest (RF) and Logistic Regression (LR) to enhance the accuracy of abnormal blood pressure prediction. The dataset, obtained from Kaggle, includes various clinical and lifestyle-related features. Data preprocessing involved handling missing values, label encoding, and normalization of numerical features. Evaluation of individual models showed that Random Forest achieved an accuracy of 0.83, while Logistic Regression reached 0.75. The sequence model, which incorporates Random Forest-generated prediction probabilities as an additional feature in Logistic Regression, improved the prediction performance with an accuracy of 0.84. Feature importance analysis identified hemoglobin level, chronic kidney disease, and genetic pedigree coefficient as the most influential predictors in classifying abnormal blood pressure. These findings highlight the effectiveness of the sequence approach in addressing the complexity of medical data and improving the precision of clinical decision support systems for hypertension diagnosis and management. Recommendations include developing advanced ensemble models, collecting longitudinal data, and conducting external validation to enhance model generalizability across diverse clinical populations.