Detection Of Variables For The Diagnosis Of Overweight And Obesity In Young Chileans Using Machine Learning Techniques.

Overweight and obesity are considered epidemic problems. The number of factors involved in developing extra body fat makes harder the detection of this problem. Therefore, among the several variables and their levels presented in overweight and obese people, there is a need to improve the classification of people with these conditions. To this aim, in this paper, we conducted a variable analysis from biochemical and lipid profiles in young Chileans with normal weight, overweight, and obesity using machine learning techniques. XGBoost library was selected as the classifier. 21 variables (13 from biochemical and 8 from lipid profiles) were chosen as features. 100 iterations were conducted, and an 80% cross-validation was obtained. The variables with greater relevance in the classification task were total cholesterol, glycemia, LDH enzyme, bilirubin, and VLDL cholesterol. All of these, except bilirubin, are consistent with previous research in which these features have been used to assess risk factors for developing overweight or obesity. Then, further research must include a deep study regarding bilirubin's influence over these conditions.