Inspiration

Our school is located in the Bronx, NY. As new drivers, we were taken aback by the appalling lack of road etiquette.

What it does

We take in inputs of the borough and time of day and then using a neural network trained by historical data from the New York Department of Motor Vehicles and New York City open data, we predict the probability of a collision at that time and in that borough.

How we built it

The function consists of a hidden value which self-adjusts over each iteration.

Challenges we ran into

Initially written in C# we had to transcribe it to JavaScript to make it easier to interact with on a website.

Accomplishments that we're proud of

We built a neural network.

What we learned

Domain nameservers take 24 hours to propagate. The floor is uncomfortable to sleep on. JavaScript is difficult to teach ourselves in 24 hours.

What's next for CrashNet

We will extend the inputs past the initial two we used (i.e. include vehicle type, weather, age of driver, etc). Instead of using Particle Swarm Optimization, we would like to switch it over to back-propagation, and instead of the softmax function, we would like to use the sigmoid function.

Share this project:

Updates