A number of industry publications and media sources reported that the growth of dollar stores is driving grocery stores out of business and perpetuating food deserts. City centers and rural areas in which residents already lack access to healthy food are targets for dollar stores.
The hypothesis our team set out to test is whether dollar stores are targeting growth in food desert regions and driving out grocery stores through competition.

We used the Census business NAICS codes to gather information about population, dollar stores and grocery on a county level. We also gathered food desert ratings of individual counties from information tracked by the US Department of Agriculture.
Scikit Learn was a machine learning library applied to the study for multiple reasons. One, we wanted to see if there were correlations between population and store count by feeding the model these metrics along with employee count and earnings. Two, once a model was developed and tested, we used it to retroactively fill in gap years during which the USDA did not calculate food desert levels.

The patterns reveal that the effect of dollar store growth on number of grocery stores and ratings of food desert is concentrated in counties in which 20% to 50% of the population lacks food access. (This was shown by the range in which the machine learning results aggregated. The model was not as accurate and often dropped outlier cases.) Scaling the business data and tying it to the population before adding to the machine learning model was essential to telling a story of how economics affect people.
Our finding shows that some of the counties that are protesting the growth of dollar stores are correct to be concern about the health of local businesses and of people. Dollar stores growth are not affect wealthy communities as much as they are putting the control of food access in the hands of private corporations in at risk communities.