Wildfire Scale Prediction Using Machine Learning

Regression and neural networks on historical wildfire data

  • Modeled wildfire scale using historical meteorological and temporal features.
  • Compared linear/tree-based regressors with shallow neural networks; conducted ablations on feature subsets.
  • Performed error analysis, calibration checks, and residual diagnostics to refine generalization.

Repo: GitHub

Highlights

  • Models: Linear regression, RandomForest/XGBoost, shallow NNs
  • Pipeline: Feature engineering, scaling, CV, hyperparameter search
  • Metrics: MAE, RMSE, R²; calibration and error slicing

References