We were inspired to start this project after hearing about the Fannie Mae Data Visualization challenge, which tasked us to visualize acquisition and performance data of loans in the housing market. We wanted to take it one step further by using Tensorflow to create models of data, so we can predict similar loan behavior across different periods of time.
What it does
This project includes a website that allows for users to interact with a model that spans across the US that predicts how successful a loan will be during a given time period. It also displays graphs and visual data to better inform users of relationships that exist between acquisition and performances of loans.
How we built it
Challenges we ran into
The biggest challenges that we ran into during this project were the training of the Tensorflow model as well as the integration of the ArcGIS API into our analyzed data.
Accomplishments that we're proud of
We are proud of the model that we created to predict the performance of a loan in a given time period. This was the first time that we built a Tensorflow model to train on financial data, and we were able to generate a model that has 94% accuracy. We also integrated ArcGIS which allowed us to display the information more intuitively.
What we learned
Prior to this experience, none of us were familiar with data science studies. However, we learned a lot in this process, from becoming familiar with .csv files to using Google Colaboratory to using Tensorflow to train models.
What's next for FannieDisplae
We want to create a more intuitive and user friendly user interface with the integration of even more APIs in order to better represent our data. We also want to look at the relationship between variables within acquisition sets and performance sets and explore edge cases like natural disasters which would cause unusual loan behavior.