Heatin-Cali

Heatin-Cali is a real-time web application that uses machine learning to predict forest fires.

Comment 2

Heatin-Cali by River

Inspiration

We were very concerned with wildfires in LA and their impact on communities.
Our project was inspired by the need to help Californian residents mitigate the effects of wildfires. We focused on three key areas:

  • Rescue Efforts – Providing timely risk assessments to aid in emergency response.
  • Containment and Risk Management – Helping officials and residents plan proactive measures.

This led to the creation of Heatin-Cali by River, a website that provides real-time wildfire risk assessments at the county level across California.

What it does

Heatin-Cali displays a heatmap indicating wildfire risk levels across California counties. The risk is categorized into four levels: Safe, Caution, Warning, and Danger, helping users quickly assess potential threats.

How we built it

We built the platform using Next.js for the frontend and Flask for the backend.

  • We trained a wildfire likelihood prediction model using Random Forest.
  • Initially, we tried Linear Regression, but it provided only ~20% accuracy.
  • Switching to Random Forest significantly improved accuracy to ~78% on our test dataset.

Challenges we ran into

One of our biggest challenges was deployment:

  • Vercel did not support our model because scikit-learn was too large to deploy.
  • We attempted to convert the model to ONNX, but while it was smaller, its accuracy dropped below 10%.
  • Eventually, we decided to host the Flask API separately on Render while keeping our Next.js frontend on Render as well.

Accomplishments that we're proud of

We are proud of our persistence and teamwork, overcoming multiple technical roadblocks.

  • We adapted quickly when our initial model failed.
  • We successfully deployed a working wildfire prediction tool despite infrastructure limitations.
  • We supported each other throughout the process, learning and improving along the way.

What we learned

Throughout this project, we learned how to:

  • Collaborate effectively across different technical domains.
  • Iterate on machine learning models to improve accuracy.
  • Deploy applications under real-world constraints and adjust for scalability.

What's next for Heatin-Cali

We plan to expand the platform by:

  • Extending wildfire predictions from 1 day to a 5-day forecast.
  • Integrating real-time alerts and notifications to send text messages to users when wildfire risks escalate.
  • Enhancing our model's accuracy by incorporating more environmental and historical data sources.

This is just the beginning, and we’re excited to keep improving Heatin-Cali to better serve Californian communities.

Built With

Share this project:

Updates