USING BIG DATA AND COMPUTER LEARNING IN THE LAB

Hi! My name is Oliver Allen, but most people call me Mel. I’m a second year student studying Computer Science and Math with a minor in Art History, and I also play marimba when I’m not studying or writing programs. I started working at the Critical Food Studies Lab in January of 2019 doing basic statistics, but now I am focused on applying my skills in computer science and machine learning to analyze data surrounding food justice and sustainability.
              I started the semester by working with Lab Director Angela Babb to create a dataset of average food consumption and nutrient profiles to assist in analysis of the Thrifty Food Plan calculation. The Thrifty Food Plan is the basis of how the USDA allocates Supplemental Nutrition Assistance Program (SNAP) benefits to over 42 million people annually, but the latest released calculation was done with data from 2001, which is nearly 20 years old. In order to perform a relevant analysis of the Thrifty Food Plan calculation we needed an updated dataset of food consumption and relevant nutrient profiles for each of the foods in the USDA Dietary Guidelines for Americans. I had to approximate the dataset used by the USDA with something built from publically available data, so I ended up downloading all of the responses to the National Health and Nutrition Examination Survey (NHANES) for 2015-2016 and working with the data in R to separate it into each of the age-sex groups specified by the Thrifty Food Plan. Each row was a food group (for example provolone cheese, roast beef, Dippin’ Dots, etc.) and there were two columns representing national consumption data and then information on the food’s nutrient profiles. Eventually we had a dataset that could at least in part approximate the one used for the Thrifty Food Plan calculation and facilitate further analysis of the USDA’s model.
            In the latter half of the semester I worked to apply machine learning techniques to data about behavior surrounding switching to a sustainable diet. I used data from a survey given by the Sustainable Food Systems Science research group at IU to approximately 500 heads of households in Indiana, and I implemented a machine learning model to predict behavior around switching to a sustainable diet from demographic data. I also used this model to analyze which demographic features are most important when deciding whether a person will switch to a sustainable diet or not, or whether they will switch gradually over time versus switching quickly due to an event or crisis. I built additional models for each of these problems to further analyze the influence of each feature in the prediction, and I hope to continue looking into applications of machine learning with data around food security and sustainability to gain insight into why people make the choices they do surrounding food and diet. 

---