Predicting wildfire vulnerability using logistic regression and artificial neural networks: a case study in Brazil's Federal District