Hypertension Classification using Supervised and Semi-Supervised Machine Learning

Hypertension, a major cardiovascular risk factor, requires accurate risk stratification for effective management. This study aims to implement both supervised and semi-supervised learning to enhance risk prediction by combining labelled and unlabelled data, which is crucial due to the scarcity of fully labelled datasets. Labelled data is costly and time-consuming to obtain, requiring expert annotation and facing data privacy concerns. As a result, many medical data remains unlabelled. By incorporating labelled and unlabelled data through semi-supervised techniques, we develop robust risk stratification tools for personalized interventions and better hypertension management. Six machine-learning models: k-Nearest Neighbour, Logistic Regression, Naive Bayes, Random Forest, Support Vector Machine, and Artificial Neural Network are trained and tested using the public dataset in MATLAB. The performance of each algorithm is analysed based on accuracy, precision, recall, AUC, and specificity. Results indicate that in supervised learning, Random Forest achieved the highest accuracy at 99.52%, followed by SVM at 99.09% and ANN at 97.27%. In semi-supervised learning, Random Forest also outperformed the other models across all performance metrics.