| As one of the forest fire-prone regions in China, Anhui Province requires improved accuracy in fire prediction and early warning. This study developed an hourly-scale forest fire prediction model based on multidimensional influencing factors, including meteorological, topographic, and vegetation data. In this study, six machine learning algorithms, namely Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Gradient Boosting Machine (LightGBM), and CatBoost, were selected for fire risk prediction modeling. The model performance was compared and analyzed. The results showed that the XGBoost model achieved a prediction accuracy of over 95% on both the training set and the test set. In actual prediction data, the accuracy exceeded 82%, significantly outperforming other models. It is suitable for precise prediction of forest fire risks in Anhui Province. Feature importance analysis using the optimal XGBoost model revealed that vegetation indicators play a crucial role in fire prediction. Based on the prediction results, spatial and seasonal distribution maps of forest fire risk in Anhui Province for the year 2020 were generated, demonstrating clear seasonal and regional variability. This study is the first to implement an hourly-scale forest fire prediction model based on meteorological data across Anhui Province using machine learning techniques. The findings confirm the feasibility and effectiveness of integrating multi-source data with intelligent algorithms for forest fire prediction, providing a scientific basis for risk assessment and precision prevention in the region. Furthermore, it offers technical support and methodological reference for enhancing forest ecological security and disaster management in China. |
