Imagine a world where the number of mass shootings in the U.S. per year don't align with the number of days. With the recent Thousand Oaks shooting, we wanted to make something that would accurately predict the probability that a place has a mass shooting given a zipcode and future date.

What it does

When you type in a zipcode, the corresponding city is queried in the prediction results of our neural network in order to get a probability. This probability is scaled accordingly and represented as a red circle of varying size on our U.S. map. We also made a donation link that takes in credit card information and processes it.

How we built it

We trained our neural network with datasets on gun violence in various cities. We did a ton of dataset cleaning in order to find just what we needed, and trained our network using scikit-learn. We also used the stdlib api in order to pass data around so that the input zipcode could be sent to the right place, and we also used the Stripe api to handle credit card donation transactions. We used d3.js and other external topological javascript libraries in order to create a map of the U.S. that could be decorated. We then put it all together with some javascript, HTML and CSS.

Challenges we ran into

We had lots of challenges with this project. d3.js was hard to jump right into, as it is such a huge library that correlates data with visualization. Cleaning the data was challenging as well, because people tend not to think twice before throwing data into a big csv. Sending data around files without the usage of a server was challenging, and we managed to bypass that with the stdlib api.

Accomplishments that we're proud of

A trained neural network that predicts the probability of a mass shooting given a zipcode. A beautiful topological map of the United States in d3. Integration of microservices through APIs we had never used before.

What we learned

Doing new things is hard, but ultimately worthwhile!

What's next for Ceasefire

We will be working on a better, real-time mapping of mass shootings data. We will also need to improve our neural network by tidying up our data more.

Share this project: