Fariborz Rahimi

In this paper, we determined the most important determinants of the bio-impedance phase angle in a subpopulation of patients suffer from cardiovascular disease. A total of 100 participants who were diagnosed as cardiovascular disease patients, participated in this study. Body composition measurements were collected from 51 males and 49 females using TANITA body composition analyzer. Four classes of machine learning algorithms were carried out to build a predictive model for predicting the accurate values of the phase angle. Fourteen initial features from the subjects’ body including weight, height, sex, age, fat mass, fat-free mass, bone mass, muscle mass, body mass index, total body water, intracellular and extracellular body water, basal metabolic rate, and visceral fat rate were used for this model. Feature importance extraction was separately performed for each class of algorithms to investigate the most effective determinants associated with phase angle variation. Performing different classes of machine learning regression models including the Linear Regression, Support vector regressions (SVR), Regression Trees and Ensemble Learning, along with comparing the obtained values of Root Mean Squared Error (RMSE), Mean Squared Error (MSE), and Mean Absolute Error (MAE), it can be indicated that the SVR method with linear kernel performs the better prediction results with error measurement of RMSE = 0.612. Results have shown that, by analyzing the importance of the features, the intracellular water and sex have the highest impact on the phase angle variations in patients with cardiovascular disease followed by total body water, basal metabolic rate, and age.