Early Detection of Hypertension Using Explainable AI Model: Random Forest Model

Hypertension is a leading cause of mortality worldwide. Early diagnosis and effective risk stratification are essential to reduce the prevalence of hypertension. The primary aim was to develop and evaluate a robust predictive model for assessing the individual risk of a binary outcome (e.g., hypertension) using Random Forest classification. The problem addressed stems from the increasing need for accurate and interpretable models in binary classification tasks, especially in health-related fields, where early prediction can inform timely interventions. A total of 4187 samples were partitioned into a training set (70%, n= 2930) and a test set (30%, n= 1257) to ensure a balanced representation of the target variable. Random Forest model with tuned hyperparameters achieved impressive performance. The Random Forest model was trained by varying the mtry values from 2 to 12. The best-performing model had an mtry of 2, achieving an AUC (ROC) of 0.9524, sensitivity of 0.9146 (SD= 0.0159), and specificity of 0.8570 (SD= 0.0454). The random Forest model demonstrated strong potential for binary classification tasks, offering both a high discriminative ability and reliable performance across metrics. This model is recommended for adoption in predictive analytics frameworks in healthcare and other high-stake decision environments. Policy implications include leveraging machine learning tools such as Random Forest to support early identification and data-driven intervention strategies.