ABSTRACT
The United States currently is the most obese country in the world, with over a third of the population weighing in well above healthy rates. Obesity also causes a slew of other health problems, including diabetes. Americans should to be able to know their risk for these potentially life-threatening conditions, and what features lead to a higher risk for these prevalent health problems. Using different machine learners, such as neural nets and decision trees trained on the food availability data from the United States Department of Agriculture, we can help overweight Americans asses their risk for further health issues.
Our research led us to a few interesting conclusions. Neural nets performed the best for this prediction task with very high correlation for obesity and diabetes on the test set, 0.7665 and 0.9001 respectively. Decision trees also performed quite well on the test set with correlations of 0.7667 and 0.872. The most important factors to predicting obesity and diabetes rates were poverty rates, average household income, and percentage of students eligible for free lunch, as seen in the strong correlations between these factors and obesity and diabetes rates in Figure 1 and Figure 2, respectively. As these factors are all strongly related to financial status of a community as a whole, it brings into question bigger problems about the correlation between poverty, effective government assistance, and making healthy options more accessible.
Our research led us to a few interesting conclusions. Neural nets performed the best for this prediction task with very high correlation for obesity and diabetes on the test set, 0.7665 and 0.9001 respectively. Decision trees also performed quite well on the test set with correlations of 0.7667 and 0.872. The most important factors to predicting obesity and diabetes rates were poverty rates, average household income, and percentage of students eligible for free lunch, as seen in the strong correlations between these factors and obesity and diabetes rates in Figure 1 and Figure 2, respectively. As these factors are all strongly related to financial status of a community as a whole, it brings into question bigger problems about the correlation between poverty, effective government assistance, and making healthy options more accessible.