Investigating Risk Factors Affecting Obesity Rates in the United States based on Artificial Neural Network and Partial Least Squares Regression Models

Background/Purpose: Recently, obesity rates across the United States have been on a continuous incline. The morbid chronic disorders associated with this disease correlate to those of the leading causes of death. With younger generations engaging in this epidemic, the factors affecting obesity should be unveiled in order to interrupt this trend. This study attempts to identify obesity- related risk factors as well as determine the strength and direction of these associations across the United States using the Artificial Neural Network (ANN) and Partial Least Squares (PLS) Regression models. Method: ANN model tested and trained data was collected by the U.S. Department of Labor for years 2000- 2006. This model was used to rank the relative importance of suggested risk factors. PLS regression validated study results. Risk factors analyzed for possible correlation with obesity rate included: per capita income, unemployment rate, poverty rate, lack/ possession of high school diploma, alcohol consumption, smoking habits, physical exercise, violent crime rate, etc. Results: Of the 14 variables analyzed, 10 were statistically associated (α= 0.01). The factors that had the highest impact on the obesity trends were: per capita income, cost of physician care, alcohol consumption, year, cost of hospital care, and cigarette smoking habits. Implications: With obesity increasing among adults and children, determining the factors associated with obesity rates will allow for the development of programs/ interventions that can stall or decrease this trend. As a result, solutions to this problem will impact the chronic disorders that are acquired due to weight.