FireFinder

Inspiration

Wildfires are inherently a big issue in California with dozens happening each year. Being students in California with our friend almost caught in a wildfire near lake Tahoe, we figured that a predictor for authorities to be able to find wildfires before they come would be a solution to help save more lives.

What it does

FireFinder uses a dataset for Californian wildfires from 2013 - 2020 to predict the occurences of the next wildires to come.

How we built it

We collected a California Wildfire dataset from Kaggle with the latitude and longitude. Afterwards we used those coordinates and time of the wildfire to find the temperature and precipitation at the time using the VisualWeather API. Then, we trained a machine learning model (neural network) with the past rainfall data, precipitation and temperature to predict future wildfires.

Challenges we ran into

Working through the limitations of the rainfall data. There were limitations (1000 calls per day) for the API and 1600 pieces of data to match. Therefore, we had to work around this by calling the API over two days and never making a mistake the script. If that were the case, then the API would be called and no data would be returned.

Accomplishments that we're proud of

We are proud of being able to create a ML model and being able to collect all the data in time. Moreover, we are proud to be able to glue the project together and produce a working product.

What we learned

We learned how to train a ML model with Scikit-learn, and built scripts to extract data quickly. Also we learned how to put a frontend together with a backend.

Built With

• fastapi
• html5
• javascript
• kaggledatasets
• keras
• pandas
• python
• scikit-learn
• visualweatherapi