Comparing Performance of Ensemble-Based Machine Learning Algorithms to Identify Potential Obesity Risk Factors from Public Health Datasets

Societal factors such as globalization, supermarket growth, rapid unplanned urbanization, sedentary lifestyle, economical distribution, and social position gradually develop behavioral risk factors in humans. Behavioral risk factors are unhealthy habits (consumption of tobacco and alcohol), improper diet (consumption of high calorific discretionary fast foods, sweet beverages), and physical inactivity. The behavioral risks may lead to physiological risks, body–energy imbalance. Obesity is one of the foremost lifestyle diseases that leads to other health conditions, such as cardiovascular disease (CVDs), chronic obstructive pulmonary disease (COPD), cancer, diabetes type II, hypertension, and depression. It is not restricted within the boundary of age and socio-economic background. “World health organization (WHO)” has predicted that lifestyle diseases will claim 71–73% of the global death, by the end of 2020. It can be prevented with proper identification of associated risk factors and appropriate behavioral intervention plans. The key determinants of obesity are—a. age, b. weight, c. height, and d. body mass index (BMI). This paper addresses the potential of ensemble machine learning approaches to assess the associated risk factors of obesity through the evaluation of existing, publicly accessible health datasets, such as “Kaggle”, and “UCI”. Followed by, we compared our identified risk factors with the obtained risk factors from literature study. In future, we are intending to reuse the obtained knowledge to collect data from a controlled trial of adult population (age between 20 and 60) in south Norway to generate personalized, contextual, and behavioral recommendations with a smart electronic coaching (eCoaching) system for behavioral intervention for the promotion of healthy lifestyle.