Ranking of a wide multidomain set of predictor variables of children obesity by machine learning variable importance techniques

Abstract

The increased prevalence of childhood obesity is expected to translate in the near future into a concomitant soaring of multiple cardio-metabolic diseases. Obesity has a complex, multifactorial etiology, that includes multiple and multidomain potential risk factors: genetics, dietary and physical activity habits, socio-economic environment, lifestyle, etc. In addition, all these factors are expected to exert their influence through a specific and especially convoluted way during childhood, given the fast growth along this period. Machine Learning methods are the appropriate tools to model this complexity, given their ability to cope with high-dimensional, non-linear data. Here, we have analyzed by Machine Learning a sample of 221 children (6–9 years) from Madrid, Spain. Both Random Forest and Gradient Boosting Machine models have been derived to predict the body mass index from a wide set of 190 multidomain variables (including age, sex, genetic polymorphisms, lifestyle, socio-economic, diet, exercise, and gestation ones). A consensus relative importance of the predictors has been estimated through variable importance measures, implemented robustly through an iterative process that included permutation and multiple imputation. We expect this analysis will help to shed light on the most important variables associated to childhood obesity, in order to choose better treatments for its prevention.

A Survey on Machine and Deep Learning Models for Childhood and Adolescent Obesity

“…For prediction purposes, ML algorithms such as decision trees and artificial neural networks could be helpful. • Non-Parametric models Almost all of the predictor focused studies [36], [9], [37], [10], [11], [38], [39], [12], [13], [5], [40] for childhood/adolescent obesity use this category of models as predictor importance is easier to gauge. Non-parametric models such as Decision Trees (DT), k-Nearest Neighbors (KNN), and Random Forest (RF) were a popular choice for predictor-based methods [24], [26] [27], [29], [41], [34] [35].…”

“…Non-parametric models such as Decision Trees (DT), k-Nearest Neighbors (KNN), and Random Forest (RF) were a popular choice for predictor-based methods [24], [26] [27], [29], [41], [34] [35]. [36], [9], [37], [10], [11], [38], [39], [12], [13], [5], [40] , [24], [26] [27], [29], [41], [34] [35] ANN [26] [27], [31], [35] RNN…”

“…Depending on the system used appreciably different estimates of overweight and obesity with age and sex were observed in the nationally representative Colombian National Nutrition Survey of 2005. Studies have also modeled BMI percentile [9], BMI progression [9] [30], and even raw BMI [40]. In the case of raw BMI modeling, it is done when there's not much difference between the ages of the children as BMI calculation for children and adolescents is age and gender-specific.…”

“…The output variable BMI or BMI categories have been used solely to find the predictor importance. The studies on predictor ranking most commonly used Decision Tree [9] [10] [11] [12] [13] [5] and Gradient Boosting Methods (GBM) [40]. Predictor ranking models can be used for intervention programs, diagnostic and prediction tools.…”

“…Table 6 summarizes the different studies that focus on finding the most important factors for predicting overweight and obesity. Majority of the studies used datasets from United States [36] [9] [10] [11] [39] [12] [5], and the rest of the studies used datasets from Cyprus [37], Canada [38], South Korea [13], and Spain [40]. China, India, the USA, Indonesia, and Brazil are the top five countries that have been projected to have the highest number of children and adolescents with obesity by 2030.…”