基于随机森林算法构建白眉野草螟监测预警模型

为科学防治白眉野草螟Agriphila aeneociliella,以16种气象因子为自变量,以白眉野草螟发生程度为因变量,采用随机森林算法构建白眉野草螟的监测预警模型,并利用构建的模型对2010—2016年影响鲁东地区白眉野草螟发生程度的关键气象因子进行分析。结果表明,当特征值为9,决策树数量为400时,白眉野草螟监测预警模型的袋外估计错误率最低,为17.88%,轻度发生和重度发生的错误率分别为17.58%和18.18%。利用测试数据检验模型,模型错误率为20.00%。通过所构建的模型分析显示影响鲁东地区白眉野草螟发生程度的关键气象因子为平均水汽压、日最低气温、平均气温和日最高气温,其Gini值分别为18.82、14.84、13.67和9.30。

Establishment of monitoring and forecasting model for eastern grass veneer Agriphila eneociliella based on the random forest algorithm

In order to control eastern grass veneer Agriphila aeneociliella scientifically, the random forest algorithm was used to construct a monitoring and forecasting model, in which 16 meteorological factors were used as independent variables and the degree of occurrence of A. aeneociliella was used as dependent variable. The model was then used to analyze the key meteorological factors which affected the occurrence of A. aeneociliella in eastern Shandong from 2010 to 2016. The results showed that, when the number of eigenvalues was nine, and the number of decision trees was 400, the least estimated error rate of the out-of-bag estimation of the model was 17.88%, and the point error rate of low and severe occurrence was 17.58% and 18.18%, respectively. The error rate of the model was 20.00% when tested by the testing data. The constructed model showed that the key meteorological factors affecting the occurrence degree of A. aeneociliella were average water vapor pressure, daily minimum temperature; average temperature and daily maximum temperature. Their Gini values were 18.82, 14.84, 13.67 and 9.30, respectively.