Abstract: Forest fire prediction involves many influence factors, complex occurrence mechanism, and unstructured input data. To address these issues, a practical forest fire prediction system based on Bayesian Network（BN） is designed and implemented. The system inputs meteorological, geographical, vegetation and human activities data, builds BN model with historical forest fires records, performs probabilistic inference via the junction tree algorithm, and outputs the probability of forest fire occurrence. Experimental results on real records of forest fires in Yunnan province and comparison with the Canadian Forest Fire Weather Index System（FWI） demonstrate that the system is feasible and practical.

Key words: forest fire, Bayesian Network（BN）, parameter learning, junction tree algorithm, probabilistic inference

摘要： 针对林火预测具有影响因素多、机制复杂、难以结构化等特点，设计并实现了一个基于贝叶斯网络的实用林火概率预测系统。该系统以气象、植被、地理、人类活动等数据作为输入，综合林火历史数据建立贝叶斯网络模型，并应用联合树算法进行概率推理，进而预测出林火发生概率。在某省实际林火历史数据上对系统进行了测试，比较了所设计系统与加拿大火险天气指标系统（FWI）的预测性能，验证了系统的可行性和实用性。

关键词: 森林火灾, 贝叶斯网络, 参数学习, 联合树算法, 概率推理