An Intelligent Wildfire Prediction Framework Using Cellular Automata and Random Forest Regression

Abstract

Forest fires are one of the most devastating natural disasters that can significantly impact ecosystems, biodiversity, human settlements, and the environment. The increased incidence of wildfire, caused by climate change, ongoing drought and human activity, necessitates the need for accurate and reliable fire forecast systems. This study presents a Forest Fire forecast Model Based on Cellular Automata and Machine Learning to improve the early detection and prediction of forest fires in terms of the occurrence and burned area. The proposed system involves forecasting fire risk using the meteorological and environmental factors using Random Forest Regressor (RFR) algorithm and geographic spread of fire using the Cellular Automata (CA) model. The model makes use of historical forest fire data that includes characteristics like temperature, relative humidity, wind speed, rainfall, Drought Code (DC), Duff Moisture Code (DMC), Fine Fuel Moisture Code (FFMC), and Initial Spread Index (ISI). To improve the quality of data, various data preprocessing techniques like data cleaning, selection and normalization are applied before the model is trained. The performance of the proposed model Random Forest Regressor is compared with the existing model Gradient boosting regressor by using the standard criteria for regression analysis. Experimental results show that the proposed approach can effectively model nonlinear relationships between environmental factors, while providing improved predictive accuracy, generalization, and reduced overfitting and computational time. A Django-based Web application is built to allow the submission of datasets, training of models, predictions, and viewing of the results through an interactive user interface. The proposed system will enable the forest management authorities to make decisions, early warning of wildfire, minimize the environmental damage and financial losses.