Demographics and obesity using a machine learning approach: Iranian National Obesity Registry (IRNOR)
Mina Nosrati, Niloufar Abdollahpour, Mahsa Tousi, Fatemeh Shahabi, Maryam Arabi, Mahdieh Zarif, Zahra Abasalti, Fatemeh Maghsoudi, Nafiseh Hosseini, Gordon A Ferns, Khalil Kimiafar, Majid Ghayour MobarhanBACKGROUND:
Obesity and overweight represent significant global health problems. The prevalence of obesity is increasing rapidly in developing countries and is involved in the etiology of cardiovascular disease, diabetes mellitus, and several cancers. In this study, we aimed to evaluate the relationship between demographic factors and obesity using machine learning (ML) within the population identified by the IRanian National Obesity Registry (IRNOR).
MATERIALS AND METHOD:
This population-based cross-sectional study was started in 2021. A total of 2324 overweight and obese individuals were registered in IRanian National Obesity Registry (IRNOR) until October of 2022. The demographic data and anthropometric measurements of all participants were recorded. Multivariable logistic regression was used to evaluate relationships between demographic factors and obesity. Support vector machine (SVM) was applied to show the classification of important factors.
RESULTS:
Data for age, education, smoking, marital, job, medical history, physical activity, and body mass index (BMI) were recorded for 2324 individuals. Of these individuals, 71.4% were female. The mean age and BMI were 42.50 years and 32.63 kg/m 2 , respectively. In multivariable regression analysis, female gender, higher education, and walking and moderate activities were associated with lower odds and being housewife had higher odds for obesity in total population. According to the SVM model, job, marital status, sex, and smoking were the most important factors associated with obesity.
CONCLUSION:
The SVM model may be effective in detection of priorities and could be helpful for more planned investigations in this field.