Classification of Impaired Waist to Height Ratio and Waist to Hip Ratio Using Support Vector Machine

The obesity epidemic has reached a high prevalence in adults, adolescents, and children. Overweight and obesity, together with a sedentary lifestyle and family history of cardiovascular disease, anticipate a high prevalence of metabolic diseases such as metabolic syndrome (MS), insulin resistance (IR), atherosclerosis, and glucose intolerance, increasing the risk of type 2 diabetes and cardiovascular disease (CVD). Although waist circumference (WC) is one of the best predictors of CVD, IR, and MS, this measure has limits because diagnostic cut-off points vary by ethnicity and race background. The waist to height ratio (WHtR) and waist to hip ratio (WHR) are suggested as better predictors because they are universal indexes that only varied because of gender. Some studies have used machine learning techniques, such as Support vector machine (SVM), clustering techniques, and random forest, in anthropometric measures such as waist circumference, hip circumference, BMI, WHtR, and WHR to evaluate the diagnosis of metabolic dysfunctions, like obesity, insulin resistance, among others. This work aims to classified impaired WHtR and WHR subjects using anthropometric parameters and the SVM technique as a classifier. This study used a database of 1978 subjects with 26 anthropometrics variables. Results showed that the SVM performed as an acceptable classification of subjects with abnormal WHtR values and abnormal WHR values using anthropometric measurements of skinfolds and circumferences.