摘要
为了预测森林火灾发生的可能性,为森林火灾的预防预报提供依据,减少火灾损失,在Python平台上应用随机森林算法,以西班牙东北Tr’as-os-Montes地区Montesinho森林公园2000年1月至2003年12月的数据记录,对影响森林火灾的指标变量进行分析和评价。结果表明,随机森林算法对森林火灾预测的准确度约为80%,表明随机森林算法对森林火灾具有较好的预测能力,可用于对森林火灾的预测预报。
In order to predict the possibility of the forest fire and provide a basis for the forest fire prevention and reduce the fire loss,the present paper intends to apply a random forest algorithm on the Python platform based on the data recorded from January,2000,to December,2003,in the Montesinho forest park in the Tr’as-os-Montes region of the north-east area of Spain.The indicator variables affecting the forest fires have been analyzed and evaluated by using a series of decision-making trees through an integrated classifier of the random forest,in which each tree in the forest can help to predict the testing data independently according to its own feature.At the same time,the Bootstrapping method has been used to register the randomly chosen different samples of data with a place-back sequence.The correlation of the samples can be reduced and properly determined by using the different sampling data to train a decision-making tree plus the random selection of the index variables of nodes.By artificially adding noise to the data of the index variable and calculating the out-of-bag error,a state can be gained that,the bigger the change of the out-of-bag error,the greater the importance of the index would be.And,finally,the eventual prediction can be made by the majority vote and deciding method,i.e.the minority has to be subordinate to the majority.The results of decision show that the accuracy of the forest fire prediction by random forest algorithm can be so accurate as to about 80%,suggesting that the random forest algorithm has better prediction power for the forest fire occurrence and control.Meanwhile,the per-month ground environment temperature,the relative humidity and DMC can also be the chief driving factors of the forest fire,of which relative humidity may have the greatest impact on the probability of the forest fire.For instance,when the relative humidity is below 30%,the probability of forest fire should be the greatest.Therefore,the random forest algorithm can be used to predict the forest fire and p
作者
贾南
陈悦
康可霖
李俊锋
JIA Nan;CHEN Yue;KANG Ke-lin;LI Jun-feng(Department of Foundation,China People1 s Police University,Langfang 065000,Hebei,China;Department of the Graduate,China People's Police University,Langfang 065000,Hebei,China)
出处
《安全与环境学报》
CAS
CSCD
北大核心
2020年第4期1236-1240,共5页
Journal of Safety and Environment
基金
河北省科技计划项目(16215416)
河北省高等学校科学技术研究项目(Z2018020)。
