Hypertension Prediction in Adolescents Using Anthropometric Measurements: Do Machine Learning Models Perform Equally Well?

The use of anthropometric measurements in machine learning algorithms for hypertension prediction enables the development simple, non-invasive models. However, different were utilized conjunction with various data, either alone or combination other biophysical and lifestyle variables. It is essential to assess impacts chosen models using simple measurements. We developed tested 13 methods neural network, ensemble, classical categories predict adolescents only imbalanced dataset 2461 samples 30.1% subjects was first partitioned into 90% training 10% validation. reduced eight measurements: age, C index, ethnicity, gender, height, location, parental hypertension, waist circumference correlation coefficient. Synthetic Minority Oversampling Technique (SMOTE) combined random under-sampling used balance dataset. optimal hyperparameters assessed accuracy, precision, sensitivity, specificity, F1-score, misclassification rate, AUC on testing Across all seven performance measures, no model consistently outperformed others. LightGBM best six metrics, except whereas Decision Tree worst. proposed Bayes’ Theorem models’ applicability Sarawak adolescent population, resulting top four being LightGBM, Random Forest, XGBoost, CatBoost, bottom Logistic Regression, LogitBoost, SVM, Tree. This study demonstrates that choice has an effect outcomes.