Wildfires are a major issue in California

What it does

Predicts the probability of a wildfire inside 1 km (approximate) grids in California where a wildfire event has taken place in the past

How we built it

Downloaded the CA wildfires dataset from their website. Cleaned the dataset. Implemented our lightGBM AI model in Sagemaker with train, validate and test phases. The January 2022 data is not yet available so we predicted the January 2022 wildfires with probabilities for each grid. The Big Sur fire was predicted correctly. We also added the display of the predictions to AWS Amplify Geo website.

Challenges we ran into

The cleanup of data required some thinking through. The amplify modules may have to reinstalled after a git pull.

Accomplishments that we're proud of

The results were pretty good.

What we learned

Feature engineering is very important.

What's next for California Wildfires

Add more features to the dataset.

Share this project: