Multiclass Classification of Hypertension Severity Using Deep Learning: Modeling and Performance Evaluation of a Multilayer Perceptron Model

Hypertension is a chronic condition characterized by sustained high blood pressure, posing significant risks for cardiovascular and renal diseases. This research explores the application of machine learning using a Multilayer Perceptron (MLP) algorithm to classify hypertension severity based on anonymized electronic medical records from Al-Ihsan Regional Hospital Bandung, covering the period from January to December 2024. The dataset includes features such as age, gender, systolic and diastolic blood pressure, heart rate, history of hypertension, dizziness, and comorbidity. The model was trained and evaluated using various data splits, with the optimal performance observed at a 70:30 training-testing ratio, achieving an accuracy of 99.1 % and a low test loss of 0.068. Hyperparameter tuning and early stopping techniques were applied to enhance model generalization. The classification report revealed high precision, recall, and F1-Scores across all hypertension severity classes. Feature importance analysis using permutation tests identified systolic and diastolic blood pressures as the most significant predictors, aligning with clinical insights. This research demonstrates the potential of MLP-based models for accurate multiclass hypertension severity classification, supporting early intervention and improved patient management.