Fariborz Rahimi

Phase Angle (PhA) is one of the most clinically relevant and important parameters that describes the ratio of body reactance and resistance. It is used for evaluating the nutritional status and determining the risk of various conditions such as cancer, AIDS, and many chronic diseases. The purpose of this research was to assess the most effective factors associated with prediction of PhA in two groups of healthy and depressed obese women. This was done by constructing a predictive model using machine learning multivariate regression methods to easily assess nutritional and cellular status in different subject groups. In this study, we used the TANITA body composition analyzer to collect data from 120 obese women out of which 61 suffered from depression. Fourteen different factors from the subject’s body including sex, age, height, weight, fat mass, and muscle mass was used for the prediction of PhA using machine learning methods. Two classes of multivariable regression analyses were considered. Every method with several feature selections was trained and tested to obtain the least error for PhA estimation. Then, for each of the two groups of participants the feature selection method was implemented to optimize the model. Our findings suggest that the PhA values of healthy and depressed obese women depend on several variables in their bodies. These variables are Age, Weight, FFM, VFR, TBW, and ICW for healthy obese women and Age, fat mass, BMI, TBW, and ICW for obese women with depressions.