Our goal was to find a way that financial data could be used towards a socially good purpose. We wanted to find correlations between events that impact communities and how those events ripple out to other communities in the country.
What it does
Our product uses publicly available US state financial data from the Bureau of Economic Analysis and natural disaster data from Federal Emergency Management Agency to predict how natural disasters impact the economic well being of each state. Using machine learning, we trained a predictive model to learn statewide GDP changes around the country after a natural disaster occurs. Using the trained model we can predict the expected economic impacts when different disasters occur around the country.
How we built it
We pulled publicly available financial data and government reported disaster data to build a dataset that expressed how states would be affected by natural disaters happening in other parts of the country. Using random forest regression techniques, we trained a model on a natural disaster happening in one state, and had it predict the next four quarters of GDP impact experienced by every other state in the US. We then used unity to visualize the impact of the data.
Challenges we ran into
The scale and dimensionality of the data available made it difficult to understand and apply many data points. Our final project provided the user the option to select any of 15 disasters for all 50 states, and predict the financial impacts that disaster may have on each state for the next four quarters. We successfully implemented this type of predictive model which has a very large number of inputs and outputs with a lot of time and planning.
Log in or sign up for Devpost to join the conversation.