Premium
The role of soft drinks on children obesity: a global analysis using machine learning techniques
Author(s) -
Gregori Dario,
Lorenzoni Giulia,
Soriani Nicola,
Azzolina Danila,
Berchialla Paola
Publication year - 2016
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.30.1_supplement.1155.2
Subject(s) - anthropometry , obesity , overweight , soft drink , environmental health , demography , consumption (sociology) , medicine , random forest , geography , artificial intelligence , computer science , food science , sociology , social science , chemistry
Background Among lifestyle factors, soft drinks’ consumption has been suggested as playing a role in contributing to obesity onset. Aim of present study is to evaluate the contribution of various risk factors to obesity worldwide using a class of machine learning techniques, namely random forests. Methods Data on 2640 children 6–11 years (Argentina, Brazil, France, Georgia, Germany, UK, India, Italy and Mexico) were collected on more than 90 parameters (anthropometrics, built environment, familiar socio economic status, food and activity frequency). Given that sample size is heterogeneous across countries (from 60 up to 1640 children), role played by soft drinks associated with WHO classes of OWO (overweight or Obese) vs. Normal children was estimated using Random Forests (RF), which represent a more robust way to assess the additional contribution of weight status’ predictors compared with traditional statistical methods. RF have been implemented using 150,000 bootstrap samples using Bylander's bias‐correction approach. Incremental role of soft drinks has been evaluated on top of all the other potential predictors. One‐hundred permutations per tree were run for assessing each factor's importance, using Out‐of‐Basket (OOB) classification error rate (CER). Results Despite the fact that soft drinks consumption is very different across continents, (children were consuming never 20% Europe, 28% India, 6% South America, and very often 33% Europe, 16% India and 54% South America, p<0.001), its contribution in determining children's weight status was not significant. Conclusions The analysis of the association of soft drinks on children obesity is undermined by the great heterogeneity in data. Random forest approach provided stable and informative estimates, showing that the role played by soft drink consumption is not relevant among common recognized determinants of OWO in children.