模糊数据挖掘技术支持下的林火蔓延模型选择研究

林火蔓延是一种复杂的燃烧现象，要准确预测林火行为，就需要有科学的林火蔓延模型. 该文结合广州市森林资源和森林防火的现状，利用模糊数据挖掘（FDM）技术，寻找与预测区域复杂自然条件相匹配的林火蔓延模型来预测林火行为. 其方法是:①确定影响林火行为的主导因子(可燃物水分含量、可燃物负荷量、易燃程度和坡度)；②构建火灾数据样本集，利用FDM的模式发现与预测两大功能，归纳发现3类模式，依照择近原则匹配各模式所适用的林火蔓延模型；③根据火场指挥者实时提供的火灾信息，遵照FDM最大贴近度原则，判定新样本所接近的模式，然后自动选择适合这一模式的林火蔓延模型来预测林火行为.该方法可以提高林火行为预测的准确性. 

Selecting forest fire spreading models based on the Fuzzy Data Mining technique

Forest fire spreading is of utmost complicated burning phenomenon and the scientific forest fire spreading model is crucial to predict forest fire behavior. According to the present situation of forest resources and forest fire prevention of Guangzhou City, based on the Fuzzy Data Mining (FDM) technique, thinking of the complicated environment, the suited forest fire spreading model can be selected to predict the behavior of forest fire. As to Guangzhou, the dominant influencing factors for forest fire behavior were assessed, including the water content of combustibles, burden quantity of combustibles, inflammability and slope. Therefore, the discovering and predicting functions of FDM are used in predicting the forest fire. According to the handy principle, three modes were discovered and the corresponding forest fire spreading model matches the certain style, which is induced from the fire data warehouse. Furthermore, so long as commander provides the real_time fire indices, the closely model can be ascertained by FDM. And the suited forest fire spreading model can be selected automatically to predict the forest fire behavior. By this means, the accuracy of predicting the behavior of forest fire can be improved.