Forest Fire Risk Assessment and Prediction Model Based on AHP-BP and SaDe-ELM

Under the circumstances of global warming and increasing industrialisation, the possibility of fire in forested areas is increasing, and the importance of accurately assessing and predicting the state of forest fire risk is increasing. Therefore, this paper aims to use statistical modelling and machine learning to assess and predict the forest fire risk status. This paper constructs a ‘climate-geography-human’ forest fire risk indicator system, uses hierarchical analysis to obtain the subjective weights of the indicators, and then uses BP neural network fitting to obtain the objective weights of the indicators, and finally uses the multivariate fusion assignment method to establish a multiscale-parameterized forest fire risk indicator system. Finally, the multivariate fusion assignment method is used to establish a multiscale-parametric forest fire risk assessment model, and a heat map comparing China's historical disaster situation and forest fire risk index is drawn, with good assessment results. Meanwhile, this paper optimises the Extreme Learning Machine (ELM) using the Adaptive Differential Evolutionary Algorithm (SaDe), and constructs a dynamic prediction model of forest fire risk based on machine learning. Finally, improvement measures are proposed for forest fire prevention and management with respect to the integrated weighting of each indicator.