These notebooks try to understand the factors that exacerbate wildfires (sizes). The dataset being used is provided by the Government of Alberta which gives data of over 20000 wildfires from 2006 to 2021, each with 50 describing features like location, wind, humidity, temperature, etc.

The eda.ipynb notebook goes over the data and visualizes it to provide a better understanding of how the features relate to determining the size and severity of the wildfire.

The models.ipynb file creates a data processing pipeline using scikit-learn where it one-hot encodes various categorical columns, normalizes numerical values, and performs feature engineering to better extract information from the features. At the end, it uses Random Forest, XGBoost, Decision Trees (Balanced or balanced with bagging), Support Vector Machines, K-Nearest Neighbors. GridsearchCV is also used to perform cross validation, determining the best hyperparameters to use for each model.

The best of the models (XGBoost) was able to achieve a testing accuracy of 92% with an F1 score of 0.71.