基于气象因子的随机森林算法在湘中丘陵区林火预测中的应用

摘要

基于湘中丘陵区1988-2014气象观测数据和森林火灾历史数据,应用二项逻辑斯蒂回归模型和随机森林算法,将样本数据随机分成训练样本(60％)和测试样本(40％),重复5次拟合,将每次拟合中筛选出的显著特征变量组成全样本数据进行拟合及交叉验证,建立湖南水口山地区林火发生预测模型.结果表明,林火的发生与林分内日最小相对湿度、细小可燃物湿度码和干旱码显著相关;随机森林算法的预测精度在所有样本组合的模拟中均比二项逻辑斯蒂回归模型的预测精度高7％～10％,即使在交叉验证中,前者的预测精度也要高10％左右,表明随机森林算法具有一定的预测优势和现实应用价值,可用于湘中丘陵地区林火预测和决策管理.%Based on the meteorological data and historical data of forest fire between 1988 and 2014 in the hilly areas in central Hunan Province,logistic regression (LR) model and random forest (RF) algorithm were used to identify the relationship between fire occurrence and meteorological factors.Dataset was randomly divided into training (60％) and validation (40％) samples,fittings were repeated for 5 times,significant predictors which were screened out by each fitting were used to constitute complete samples to conduct fitting operation and cross validation,two methods were applied to establish fire prediction model for Shuikou mountainous area.The results indicated that daily minimum relative humidity,fine fuel moisture code (FFMC) and drought code (DC) had significant correlation with forest fire.In the simulation of all the samples,the prediction accuracy of RF was higher 7 ％-10％ than that of LR.Even in the cross validation test,the former's prediction accuracy was also about 10％ higher.Those results revealed that the RF model could be used in the fire prediction and decision management in the hilly areas in central Hunan Province.