Environmental Fire Hazard Detection and Prediction using Random Forest Algorithm

Forest fire has become an issue that looms over the entire world, mainly because of climate change. This can be catastrophic for humans as well as animals. This paper aims at detecting and predicting potential wildfires using Random Forest Algorithm. Forest fire data from the Montesinho natural park in Portugal's Tras-os-Montes upper east zone, collected over three years, is used in this study (January 2000 to December 2003). The data used is linearly correlated and helps detect the ideal environmental conditions for a fire-Pearson's method to make predictions and raise the alarm to avoid the worst. Based on Random Forest regression, a correlation matrix was obtained with an absolute mean error of 0.66. Hence, this method can be deployed for the effective prediction of the forest fire.