A Data Mining Approach for Prediction of Forest Fire Using Machine Learning Algorithm

Every year, millions of hectares of forests are destroyed by forest fires, which are a common and significant natural calamity that seriously endangers both human life and property. For the purpose of developing quick risk management plans and putting into practice efficient firefighting techniques, precise quantitative forecasting of forest fire spread is crucial. The Forest Fire Spread Behaviour Prediction (FFSBP) model, which includes two essential components—the Forest Fire Spread Process Prediction (FFSPP) model and the Forest Fire Spread Results Prediction (FFSRP) model—is presented in this work in answer to this requirement. Cellular Automata and Wang Zhengfei models are combined to create the FFSPP model, which predicts the direction and rate of forest fire propagation. The FFSRP model, on the other hand, employs machine learning techniques to anticipate the size of the burnt area. An actual case study is conducted utilising the "3.29 Forest Fire" disaster in China to verify the effectiveness of the suggested models. This instance is used to verify the FFSPP model, while an actual fire dataset from Portugal's Montesinho National Forest Park is used to evaluate the FFSRP model. The validation process's findings show that the FFSPP model forecasts a burnt area of 286.81 hm2 with a corresponding relative error of 28.94% during the "3.29 Forest Fire's" natural growth phase. Compared to the Farsite and Prometheus fire behaviour simulation models, this relative inaccuracy is noticeably lower. Furthermore, the FFSRP model has excellent prediction performance, especially when considering small and medium-sized fire situations. These results highlight the FFSBP model's potential as a useful instrument for improving the accuracy of forest fire predictions and enabling more effective risk reduction and suppression tactics.