Distributed random forest for predicting forest wildfires based on weather data

Abstract

Forest fires pose a significant threat to ecosystems, economies, and human settlements. Accurate prediction of forest fires can aid in timely interventions, resource allocation, and effective management strategies. This study aimed to develop a machine learning model to predict forest fire occurrences based on various environmental and meteorological variables. Using a dataset comprising variables such as temperature, humidity, wind speed, and moisture codes (FFMC, DMC, DC, and ISI), we employed Distributed Random Forest (DRF) and a 5-fold cross-validation approach on training data to assess the model’s performance. The model demonstrated high discriminatory power with an AUC of 0.989 and a low Mean Squared Error (MSE) of 0.041. The results underscored the critical role of weather conditions and fuel moisture content in influencing fire occurrences. The study’s findings have implications for forest management, emphasizing the potential of machine learning in shaping fire prevention strategies and safeguarding forest ecosystems.