Enhancing Hypertension Risk Assessment Through Machine Learning Models

¹ School of Technology, Woxsen University, Kamkole, Sadasivpet, Hyderabad - 502345, India
² School of Sciences, Woxsen University, Kamkole, Sadasivpet, Hyderabad - 502345, India
³ Electrical Engineering, Electronics, and Automatic Control, Universität de Girona, Campus de Montilivi, s/n 17071, Girona.

Published online: 19 June 2026

Abstract

Hypertension is also a significant health issue in the world that needs precise early risk prediction mechanisms. The present study suggests an improved ensemble-based machine learning model combining XGBoost, LightGBM, and CatBoost classifiers with the Hyperparameter optimization of Optuna. The pipeline consists of correction of imbalance by the use of SMOTE, structural preprocessing, and feature engineering on the basis of clinical knowledge to enhance the minority-class detection. The optimized CatBoost has experimental results of 0.8852 F1-score and the final soft-voting ensemble has 0.8986 accuracy and 0.9000 weighted F1-score. The suggested framework offers better sensitivity in the balance of classes and provides computationally efficient hypertension risk prediction clinically meaningful enough to be efficiently integrated in decision-supporting.