As a team we are all enamored with the idea of building something to help the community. With this combined interest, we explored datasets that affected a large number of people in a meaningful way. We found that many of the datasets on homelessness had a lot of potential, so we decided to focus on predictive models using that data.
What it does
The model uses population, economic, and demographic data in order to predict the unsheltered homeless population in regions across the United States.
How we built it
We compiled and cleaned data from several sources then applied multiple modeling techniques in python in order to maximize accuracy.
Challenges we ran into
We ran into issues combining features from multiple sources without leaving empty spaces. Additionally, we had trouble calculating accuracy of a neural network with cross validation.
Accomplishments that we're proud of
We are proud of the way we structured our problem solving process in order to develop a robust model in a short period of time. We are mainly proud of the research and planning that went into this project.
What we learned
We learned to visualize our data before starting a project, so that we don't miss any easy solutions that might be right in front of us.
What's next for Predicting Homelessness
We would like to introduce a feature that could be manipulated (such as the aid dispersal by location), so that decisional impact can be predicted.