Intelligent fire risk assessment and emergency decision support system based on big data and machine learning

This paper proposes an intelligent fire risk assessment and emergency decision support system based on big data and machine learning. The system collects multi-source fire data through Internet of Things devices, and uses big data technology for cleaning, fusion and storage, thus building a huge data resource pool. In the aspect of risk assessment, the improved Stacking integrated learning framework is adopted, combining XGBoost, LightGBM and GraphConv algorithms, and the weight coefficient is determined by Bayesian optimization. At the same time, the spatio-temporal propagation model is introduced to simulate the dynamic changes of fire risk, which significantly improves the accuracy and robustness of fire risk prediction. The emergency decision support layer optimizes the evacuation path planning through the improved A* algorithm, designs a multi-objective optimization model to realize the efficient scheduling of rescue resources, and uses NSGA-II algorithm to solve the Pareto frontier, providing decision makers with a variety of trade-off schemes. The system is based on high-performance hardware cluster and distributed software framework, and deployed by micro-service architecture, which supports high concurrent data processing and real-time decision response. The performance test results show that the system meets the timeliness requirements of fire emergency scenarios in terms of data processing speed, risk assessment delay and decision response time, and shows good stability and robustness under extreme conditions such as high concurrent writing, network jitter and single-node failure. The system provides scientific fire risk assessment and emergency decision support for the fire department, and effectively improves the intelligent level of fire prevention and control.