Hypertension Detection based on Machine Learning

The paper presents the occurrence of high blood pressure for three categories of persons, namely the normal population, patients with hypertension, and pregnant women with or without hypertension and preeclampsia in the context of a proposed healthcare system. The input data is public and it is provided by the Massachusetts University Amherst and National Health and Nutrition Survey. The data have been processed using four machine learning classifiers, namely Gaussian Naive Bayes, Logistic regression, Random Forest and Support Vector Machines. The criteria of the analysis are based on the absence or occurrence of hypertension, precision, recall, F1-score and population indicators. For the analyzed cases, the persons who did not present high blood pressure have been correctly detected, while half of the cases for which hypertension was present, proved to be true. As a conclusion, the automatic detection of healthcare parameters that exceed the allowed thresholds and are determined based on machine learning proves to be important for monitoring and prevention of critical health issues of the persons who belong to diverse categories.