Inspiration

We were moved by the amount of destruction caused by wildfires in America, most prominently on the west coast in recent years.

What it does

We use several attributes, including meteorological data (provided by an API) to create a novel model which predicts the likeliness and severity of wildfires.

How we built it

The model was built using Python primarily, with R and Stata supplementing when necessary. Within Python, we primarily used pandas, numpy, and scikit-learn. The front end was done in JavaScript, HTML, CSS, and Folium.

Our particular model used a Random-Forest ensemble algorithm. We also trained a SVM model, which was outperformed by the former.

Challenges we ran into

The machine-learning algorithms took extensive time and memory to train. Some trial-and-error was required before we decided on our final model.

Accomplishments that we're proud of

We did it!!!

What we learned

How to use scikit-learn and caret.

What's next for FireWatch

Our model would benefit from more recent data, given the rapidly changing climate conditions.

Share this project:
×

Updates