Using Simulation and SVM Methods to Determine the Relationship Between U.S. County-Level Adult Obesity Rate and Its Risk Factors

The adult obesity rate doubled within the past two decades based on data from the Center for Disease Control (CDC). Simulation and support vector machine (SVM) models were conducted to determine the relationship between U.S. county-level adult obesity rate and its risk factors. Simulation model was run on the initial analysis for multiple linear regression function. The SVM model seeks an optimal hyperplane to separate data from different classes through the computational shortcut of kernel functions. The outcome variable was the adult obesity rate. All risk factors were categorized into four domains of the social ecological model: biological/behavioral factor, socioeconomic status, food environment, and physical environment. Of the 23 risk factors related to adult obesity, the most influential eight factors were identified including physical inactivity, natural amenity, median household income, and percent of all restaurants being fast food. The study results from simulation and SVM models were consistent with those in the literature. This study accomplished its objective by confirming the research hypothesis that the obesity epidemic in all U.S. counties exhibits discernable patterns affected by four domains of social ecological framework: biological or behavioral factor, socioeconomic status, food environment, and physical environment. By analyzing multiple risk factors of obesity in the communities, may lead to the proposal of more comprehensive and integrated policies and intervention programs to solve the population-based problem.