TRADITIONAL LOGISTIC REGRESSION AND MACHINE LEARNING APPROACHES OF SOCIODEMOGRAPHIC AND ANTHROPOMETRIC FACTORS INFLUENCING HYPERTENSION IN ATHLETES

Abstract

The type and intensity of exercise performed by athletes play an important role in affecting blood pressure stability, putting them at risk of developing hypertension. Hypertension, or high blood pressure, is a medical condition in which the blood pressure in the arteries rises above normal limits. Hypertension in athletes becomes an essential factor in real cases if not detected early. Therefore, this study aims to model and analyse the sociodemographic and anthropometric factors that influence the incidence of hypertension. The data used in this study are primary data from 200 athlete selection participants at the University of Surabaya and the Indonesian National Sports Committee (INSC) of East Java. This research method proposes to compare the traditional approach with machine learning to prove the accuracy comparison of the model's goodness, where both approaches are proposed by considering the novelty proposed through the machine learning approach but still maximizing the traditional approach. The proposed methods are binary logistic regression, binary logistic regression with the addition of random effects, highly randomized tree, and support vector classification. The binary logistic regression model is better than the binary logistic regression model with random effects, random trees, and support vector classification because the accuracy, sensitivity, specificity, and F1-score value (68.5%, 69%, 68%, and 68.8%) is highest than the others. Other results showed that the waist circumference variable, the father's occupation variable, and the salary variable significantly affected hypertension at the 5% significance level.