Wildfire Severity Prediction

Inspiration

We were inspired by the impact of the devastating wildfires in the American west in recent years. We wanted a way for users to be able to understand the risk of fire in their community.

What it does

It implements regression analysis over a dataset of 55,000+ historical wildfires to predict the severity of potential wildfires across the United States. It takes into account a location's current weather and wildfire history to give a real-time updating prediction.

How we built it

We implemented our machine learning model in Python using Scikit-Learn. We used HTML, CSS, JavaScript to build the website and use Node.js to deploy APIs and achieve backend and frontend connection.

Challenges we ran into

We were not very familiar with frontend and backend communication at first, so we self-studied it on the site and learned from the mentors.
Initially, finding a suitable dataset for the model also took we a lot of time. We had to slightly switch our topic to meet the with what we have at hand.

Accomplishments that we're proud of

We divide our team into data analysis group and web development groups. We are proud of our ability to process data and build a model off of that data.
The fit of our data is decent even when the dataset of our model is relatively biased in terms of sampling location.

What we learned

We learned how to collect user input from a dynamic dropdown menu, use APIs to get relevant weather information, give it to our model and send the information back to the frontend.

What's next for Wildfire Severity Prediction

Given current dataset, we are able to predict wildfire size mainly in the Southwest region of the United States. The next step for this project will be getting larger dataset that can train our model better. With dataset that are collected on longer time span and over broader area, our model will generalize even better with the help of machine learning