This study used machine learning approaches to investigate whether body composition indices derived from bioelectrical impedance analysis (BIA) can predict hypertension in a cohort of patients. The Fasa cohort study recruited 10,000 people and assessed predisposing factors for non-communicable diseases in rural regions of Fasa, Iran. A subset of 4663 records was used, with 2156 male and 2507 female participants aged 35-70. Hypertension diagnosis was based on the blood pressure threshold defined by ACC/AHA guidelines. Body composition analysis was performed using eight electrodes, and the following variables were measured: fat percentage, BMR, FATM, FATP, FFM, LLFATP, RLFATP, LLFFM, RLFFM, LLFATM, RLFATM, LAFATP, RAFATP, LAFATM, RAFATM, LAFFM, RAFFM, TRFATP, TRFATM, and TRFFM. The target feature was the discrete binary variable of hypertension. Support Vector Classifier (SVC), Decision Tree (DT), Stochastic Gradient Descend (SGD) Classifier, Logistic Regression (LR), and Random Forest (RF) algorithms were used to analyze the data. Results showed that the SVC model had the highest accuracy of 0.842.
Related Posts
