Predictive Modeling of Wildfires in the United States

Abstract

This research utilizes wildfire records between 1911 and 2015 to train various models to predict fire size through using temperature, wind, humidity, and precipitation as features. Our results show 1) Decision Tree based Classifier outperforms both linear and ridge regression 2) Government entities can leverage our methodology to manage wildfires more efficiently, effectively, and decreasing monetary damages.