A classifier of fire types, Trained on the dataset "1.88 Million US Wildfires" from Kaggle. The classifier is Gradient Boosting of xgboost package.
A full description of the project, including explanation about the feature engineering and the performances of other models on that problem are discussed in detail in the final_report.pdf.
This project was submitted for the course Applied Competitive Lab in Data Science (67818) of the Hebrew University of Jerusalem.
Download the dataset from Kaggle as a .csv file. Also download the shapefiles from the list in the file shapefiles_sources.txt (we couldn't upload them to this repository for they were too heavy), and same them in the same directory in a subdirectory called shapefiles.
Download all the .py and .ipynb files to the same directory as the dataset. Open main.ipynb and replace the paths of train and test datasets. You can assist the function get_work_data in the utils.py file for split the full dataset to train and test.
A full description of the code's modules appeares at the end of the file final_report.pdf.