基于随机森林模型的紫薇绒蚧微环境因子排序  

Microenvironment factor permutation of Eriococcus lagerostroemiae based on random forest model

作　　者：任向辉[1,2] 余昊 杨萌[1,2] 吕栋标 刘俊杰[1] 李鹏 REN Xianghui;YU Hao;YANG Meng;L Dongbiao;LIU Junjie;LI Peng(School of Resource and Environment Science,Henan Institute of Science and Technology,Xinxiang,Henan 453003,China;Key Laboratory for Crops Pest Management of Xinxiang,Xinxiang,Henan 453003,China)

机构地区：[1]河南科技学院资源与环境学院,河南新乡453003 [2]新乡市农作物重大有害生物防控重点实验室,河南新乡453003

出　　处：《森林与环境学报》2018年第4期506-512,共7页Journal of Forest and Environment

基　　金：国家自然科学基金面上项目(31272103);国家现代农业产业技术体系建设专项(CARS-45-SYZ10);河南科技学院重点基金项目(0500136);2015年国家级大学生创新创业项目(20130515403)

摘　　要：为探讨紫薇绒蚧的发生规律,用LAI-2000植物冠层仪和电子温湿度计分别对120条紫薇一级分枝的室外冠层信息及紫薇绒蚧生存的小气候条件进行了调查,在对紫薇绒蚧的发生及其相关微环境因子做初步的灰色关联度分析基础上进行了PCA(主成分分析)排序分析,并利用R软件进行了该害虫发生与相关环境因子基于随机森林模型的分类与回归分析,分类模型预测准确度100%。随机森林分类模型与灰色关联度分析的结果均表明紫薇绒蚧的发生与叶丛温度的关联性最明显,另外该虫的发生级别与寄主枝条的冠层数据也密切相关。随机森林模型回归分析可绘制出每个环境因子不同水平对紫薇绒蚧发生的具体影响效果图,这些效果图显示叶片数少、叶丛间湿度低且枝条遮蔽度低等具宿主长势较弱的微环境因子特征的枝条该虫发生密度较高。Using LAI-2000 plant canopy analyzer and electronic temperature hygrometer for recognizing the occurrence regularity of Eriococcus lagerostroemiae,the outdoor canopy information of the 120 pests and the microclimate data around first-order branches of Lagerstroemia indica were surveyed,and the principal component analysis(PCA)permutation of the environmental factors was carried out base on the primary analysis of grey relational analysis(GRA)between this pest′s occurrence and its relative microenvironment factors.Furthermore the classification and regression analyses with random forest model(RF model)of these data were calculated using R software,and the accuracy rate of classification was 100%.Both analysis results of GRA and RF classification model showed that this pest′s occurrence had the best related to the temperature of leaves in-depth,and the occurring level of this pest associated the host plant canopy information greatly.Finally RF regression model analysis could draw the graphs which contained every insect occurring level and every environmental factor′s level concretely,and these graphs showed that this pest would be occurred more densely in the weaker branches which had some host microenvironment characteristics such as less leaves,lower humidity of leaves in-depth and less shadowy meter.

关 键 词：紫薇绒蚧 冠层 随机森林模型 微环境因子 

分 类 号：S763.35[农业科学—森林保护学]