高级检索

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于机器学习的落叶松毛虫发生面积预测模型

张文一 景天忠 严善春

张文一, 景天忠, 严善春. 基于机器学习的落叶松毛虫发生面积预测模型[J]. 北京林业大学学报, 2017, 39(1): 85-93. doi: 10.13332/j.1000-1522.20160205
引用本文: 张文一, 景天忠, 严善春. 基于机器学习的落叶松毛虫发生面积预测模型[J]. 北京林业大学学报, 2017, 39(1): 85-93. doi: 10.13332/j.1000-1522.20160205
ZHANG Wen-yi, JING Tian-zhong, YAN Shan-chun. Studies on prediction models of Dendrolimus superans occurrence area based on machine learning[J]. Journal of Beijing Forestry University, 2017, 39(1): 85-93. doi: 10.13332/j.1000-1522.20160205
Citation: ZHANG Wen-yi, JING Tian-zhong, YAN Shan-chun. Studies on prediction models of Dendrolimus superans occurrence area based on machine learning[J]. Journal of Beijing Forestry University, 2017, 39(1): 85-93. doi: 10.13332/j.1000-1522.20160205

基于机器学习的落叶松毛虫发生面积预测模型

doi: 10.13332/j.1000-1522.20160205
基金项目: 

东北林业大学学术名师支持计划 010602071

详细信息
    作者简介:

    张文一。主要研究方向:昆虫化学生态。Email:2904261860@qq.com  地址:150040  黑龙江省哈尔滨市和兴路26号东北林业大学林学院

    责任作者:

    严善春,教授,博士生导师。主要研究方向:昆虫化学生态。Email:yanshanchun@126.com  地址:同上

  • 中图分类号: S763.3

Studies on prediction models of Dendrolimus superans occurrence area based on machine learning

  • 摘要: 落叶松毛虫为我国主要害虫之一,其发生严重影响了我国林木生长和森林资源的安全。因此,及时准确地对落叶松毛虫虫害发生趋势进行预测、预报十分必要。虫害的发生受到多种因素的影响,存在复杂的非线性关系,传统的预测方法大多为基于线性的预测,导致其预测效果不够理想。本研究选取当年3月中旬的总蒸发量、上年7月上旬的平均最低气温、当年3月下旬的极端最低气温以及上年11月上旬的平均风速作为自变量,虫害发生面积作为因变量,利用多层前馈神经网络(MLFN)、广义回归神经网络(GRNN)以及支持向量机(SVM)3种机器学习算法对落叶松毛虫发生面积进行预测,并将3种方法的预测结果与传统多元线性回归预测方法相比较。结果表明,机器学习的预测效果均在很大程度上优于多元线性回归预测,并且在3种机器学习算法中,SVM模型的预测效果最好,在30%容忍度下其预测精度可以达到100%,并且该模型还有较低的RMSE值(0.077)和较短的训练时间(1 s)。这表明,机器学习可以应用于生产实际并有效预测虫害发生面积,尤其是SVM模型可以作为一种很好的虫害发生预测手段。

     

  • 图  1  MLFN原理图

    Figure  1.  Structure of the MLFN

    图  2  神经元ij之间的关系

    Figure  2.  Connection between neurons i and j

    图  3  GRNN原理图

    Figure  3.  Structure of the GRNN

    图  4  支持向量确定最优超平面的位置

    Figure  4.  Support vectors determining the position of optimal hyperplane

    图  5  MLFN预测结果

    Figure  5.  Predicting results of MLFN model

    图  6  GRNN预测结果

    Figure  6.  Predicting results of GRNN model

    图  7  SVM模型预测结果

    Figure  7.  Predicting results of SVM model

    表  1  多元线性回归系数表

    Table  1.   Coefficients of multiple regression model

    模型
    Model
    非标准化系数
    Non standardized coefficient
    标准系数
    Standard coefficient
    tP
    系数
    Coefficient
    标准误差
    Standard error
    α010.7192.4944.2970
    X10.1630.0520.5463.1510.006
    X2-0.6220.138-0.789-4.5170
    X30.1450.050.5032.8900.01
    X4-0.6110.35-0.256-1.7460.099
    注:α0为常量,X1为蒸发量,X2为平均最低气温,X3为极端最低气温,X4为平均风速。Notes: α0 means constant variable, X1 means evaporation capacity, X2 means average minimum temperature, X3 means extreme minimum temperature, X4 means average wind speed.
    下载: 导出CSV

    表  2  多元线性回归模型汇总表

    Table  2.   Summary sheet of multiple regression model

    模型
    Model
    RR2调整R2
    Adjusted R2
    标准估计的误差
    Standard estimate error
    10.8190.6710.5940.910 51
    下载: 导出CSV

    表  3  多元线性回归预测结果

    Table  3.   Predicting results of multiple regression model

    年份
    Year
    实际值/103
    hm2Actual value/103
    ha
    预测值/103
    hm2Predicting value/103 ha
    19920.8000.90
    19990.6670.49
    20000.733-0.63
    20071.0001.29
    20110.760-0.16
    20120.8670.15
    下载: 导出CSV

    表  4  ANN预测结果

    Table  4.   Predicting results of ANN

    年份
    Year
    实际值/103 hm2
    Actual value/103 ha
    MLFN模型预测值/103 hm2
    Predicting value of MLFN/103 ha
    GRNN模型预测值/103 hm2
    Predicting value of GRNN/103 ha
    19920.8000.84(Good)0.52(Bad)
    19990.6670.99(Bad)0.13(Bad)
    20000.7330.03(Bad)0.69(Good)
    20071.0001.37(Bad)1.01(Good)
    20110.7600.30(Bad)0.66(Good)
    20120.8670.74(Good)1.00(Good)
    RMSE0.400 20.256 5
    注:30%容忍度下预测结果准确为Good,不准确为Bad。下同。Notes: under the tolerance of 30%, the precision of predicting results was labeled as Good, or as Bad. The same below.
    下载: 导出CSV

    表  5  SVM预测结果

    Table  5.   Predicting results of SVM

    年份
    Year
    实际值/103 hm2
    Actual value/103 ha
    预测值/103 hm2
    Predicting value/103 ha
    19920.8000.767(Good)
    19990.6670.770(Good)
    20000.7330.691(Good)
    20071.0001.101(Good)
    20110.7600.792(Good)
    20120.8670.971(Good)
    RMSE0.077
    下载: 导出CSV

    表  6  不同预测方法的预测效果对比

    Table  6.   Comparison in predicting results among varied predicting models

    项目
    Item
    实际值/103 hm2
    Actual value/103 ha
    多元线性回归
    Multiple regression model
    MLFN模型
    MLFNmodel
    GRNN模型
    GRNNmodel
    SVM模型
    SVMmodel
    年份
    Year
    19920.8000.900.840.080.77
    19990.6670.490.990.800.77
    20000.733-0.630.030.130.69
    20071.0001.291.371.271.10
    20110.760-0.160.300.470.79
    20120.8670.150.740.600.97
    RMSE预测准确率
    Forecasting accuracy
    0.747 40.400 20.256 50.077 0
    33.33%66.67%100%
    训练时间
    Training time
    0:00:560:00:010:00:01
    下载: 导出CSV
  • [1] 王文龙.落叶松毛虫成虫种群数量监测技术研究[D].北京: 北京林业大学, 2014. http://cdmd.cnki.com.cn/Article/CDMD-10022-1014327487.htm

    WANG W L. Technology of population dynamics monitoring of Dendrolimus superans (Butler) adult[D]. Beijing: Beijing Forestry University, 2014. http://cdmd.cnki.com.cn/Article/CDMD-10022-1014327487.htm
    [2] 邹莉, 张国权, 黄建伟.东北地区落叶松毛虫研究进展[J].安徽农业科学, 2014, 42(9): 2584-2587. doi: 10.3969/j.issn.0517-6611.2014.09.024

    ZOU L, ZHANG G Q, HUANG J W. Research progress of Dendrolimus superansis in Northeast of China[J]. Journal of Anhui Agricutural Science, 2014, 42(9): 2584-2587. doi: 10.3969/j.issn.0517-6611.2014.09.024
    [3] 高宝嘉.关于森林有害生物可持续控制的思考[J].北京林业大学学报, 1999, 21(4): 112-115. doi: 10.3321/j.issn:1000-1522.1999.04.024

    GAO B J. Consideration on sustainable control of forest pests[J]. Journal of Beijing Forestry University, 1999, 21(4): 112-115. doi: 10.3321/j.issn:1000-1522.1999.04.024
    [4] 薛贤清.在马尾松毛虫测报中应用逐步回归电算方法的研究[J].林业科学, 1984, 20(1): 42-49. http://www.linyekexue.net/CN/Y1984/V20/I1/42

    XUE X Q. A study on the application of successive regression by computerization method to the forecast of Dendrolimus punctarus[J]. Scientia Silvae Sinicae, 1984, 20(1): 42-49. http://www.linyekexue.net/CN/Y1984/V20/I1/42
    [5] 梁其伟.四川省马尾松毛虫不同发生类型区的气候条件判别分析[J].数理统计与管理, 1990(1): 12-16. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=QK000002792926

    LIANG Q W. Discriminant analysis on climatic conditions of different occurrence types of Dendrolimus punctatus in Sichuan Province[J]. Journal of Applied Statistics and Management, 1990(1): 12-16. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=QK000002792926
    [6] 李天生, 赵良彪, 卢崇飞, 等.用双重筛选逐步回归法对广西钦州县松毛虫发生进行分析与预测[J].林业科学, 1985, 21(3): 247-252. http://www.linyekexue.net/EN/abstract/abstract365.shtml

    LI T S, ZHAO L B, LU C F, et al. Analysis and predicting of the pine caterpillar of Qinzhou County in Guangxi Autonomous Region by means of double selection stepwise regression method[J]. Scientia Silvae Sinicae, 1985, 21(3): 247-252. http://www.linyekexue.net/EN/abstract/abstract365.shtml
    [7] 吴敬, 洪伟, 翁少容, 等.马尾松毛虫消长预报的数量化模型[J].林业科学, 1983, 19(1): 39-45. http://www.linyekexue.net/CN/Y1983/V19/I1/39

    WU J, HONG W, WENG S R, et al. The quantitative model for forecasting the growth and decline of Dendrolimus punctatus[J]. Scientia Silvae Sinicae, 1983, 19(1): 39-45. http://www.linyekexue.net/CN/Y1983/V19/I1/39
    [8] 叶文虎, 马小明, 李天生, 等.马尾松毛虫预测预报系统的研究[J].林业科学研究, 1990, 3(5): 427-433. http://www.lykxyj.com/ch/reader/view_abstract.aspx?file_no=19900502&flag=1

    YE W H, MA X M, LI T S, et al. The population prediction of the pine caterpillar[J]. Forest Research, 1990, 3(5): 427-433. http://www.lykxyj.com/ch/reader/view_abstract.aspx?file_no=19900502&flag=1
    [9] 王淑芬, 张真, 陈亮.马尾松毛虫防治决策专家系统[J].林业科学, 1992, 28(1): 31-38. http://d.old.wanfangdata.com.cn/Conference/205444

    WANG S F, ZHANG Z, CHEN L. An expert system for the control of Dendrolimus punctatus[J]. Scientia Silvae Sinicae, 1992, 28(1): 31-38. http://d.old.wanfangdata.com.cn/Conference/205444
    [10] 王霓虹, 缪天宇, 王阿川.基于WebGIS的森林病虫害预测预报专家系统的设计与应用[J].东北林业大学学报, 2008, 36(1): 79-80, 82. doi: 10.3969/j.issn.1000-5382.2008.01.032

    WANG N H, LIAO T Y, WANG A C. Research and application of web-based forecast expert system for forest disease and insects[J]. Journal of Northeast Forestry University, 2008, 36(1): 79-80, 82. doi: 10.3969/j.issn.1000-5382.2008.01.032
    [11] 高倩.落叶松毛虫预测模型与模拟研究[D].保定: 河北农业大学, 2012. http://cdmd.cnki.com.cn/Article/CDMD-11920-1012332275.htm

    GAO Q. The study on prediction model and simulation of Dendrolimus superans[D]. Baoding: Agricultural University of Hebei, 2012. http://cdmd.cnki.com.cn/Article/CDMD-11920-1012332275.htm
    [12] 王霓虹, 李丹, 潘华.落叶松毛虫虫害防治预测咨询平台的研建[J].北京林业大学学报, 2009, 31(增刊2): 31-35. http://bjly.chinajournal.net.cn/editorc/WebPublication/paperDigest.aspx?paperID=BJLY2009S2006&isCnki=ck01

    WANG N H, LI D, PAN H. Consultation platform of prevention and forecasting for Dendrolimus superans[J]. Journal of Beijing Forestry University, 2009, 31(Suppl. 2): 31-35. http://bjly.chinajournal.net.cn/editorc/WebPublication/paperDigest.aspx?paperID=BJLY2009S2006&isCnki=ck01
    [13] 戚莹.基于模糊神经网络的森林虫害预测预报的应用研究[D].哈尔滨: 东北林业大学, 2011. http://cdmd.cnki.com.cn/Article/CDMD-10225-1011146672.htm

    QI Y. The research on forest pest forecast based on fuzzy neural network[D]. Harbin: Northeast Forestry University, 2011. http://cdmd.cnki.com.cn/Article/CDMD-10225-1011146672.htm
    [14] 闫友彪, 陈元琰.机器学习的主要策略综述[J].计算机应用研究, 2004, 21(7): 4-10. doi: 10.3969/j.issn.1001-3695.2004.07.002

    YAN Y B, CHEN Y Y. A survey on machine learning and its main strategy[J]. Application Research of Computers, 2004, 21(7): 4-10. doi: 10.3969/j.issn.1001-3695.2004.07.002
    [15] 杨淑香, 赵慧颖, 包兴华.基于人工神经网络的落叶松毛虫发生量预测模型的研究[J].中国农学通报, 2014, 30(28): 72-75. http://d.old.wanfangdata.com.cn/Periodical/zgnxtb201428014

    YANG S X, ZHAO H Y, BAO X H. A study on the forecast model of Dendrolimus superans butler occurrence based on artificial neural network[J]. Chinese Agricultural Science Bulletin, 2014, 30(28): 72-75. http://d.old.wanfangdata.com.cn/Periodical/zgnxtb201428014
    [16] 张景阳, 潘光友.多元线性回归与BP神经网络预测模型对比与运用研究[J].昆明理工大学学报(自然科学版), 2013, 38(6): 61-67. doi: 10.3969/j.issn.1007-855x.2013.06.010

    ZHANG J Y, PAN G Y. Comparison and application of multiple regression and BP neural network prediction model[J]. Journal of Kunming University of Science and Technology (Natural Science Edition), 2013, 38(6): 61-67. doi: 10.3969/j.issn.1007-855x.2013.06.010
    [17] 毛健, 赵红东, 姚婧婧.人工神经网络的发展及应用[J].电子设计工程, 2011, 19(24): 62-65. doi: 10.3969/j.issn.1674-6236.2011.24.022

    MAO J, ZHAO H D, YAO J J. Application and prospect of artificial neural network[J]. Electronic Design Engineering, 2011, 19(24): 62-65. doi: 10.3969/j.issn.1674-6236.2011.24.022
    [18] HOPFIELD J J. Artificial neural networks[J]. IEEE Circuit & Devices Magazine, 1988, 4(5): 3-10. http://d.old.wanfangdata.com.cn/Periodical/gpxygpfx201509047
    [19] YEGNANARAYANA B. Artificial neural networks[M]. New Delhi: PHI Learning Pvt. Ltd., 2009.
    [20] DAYHOFF J E, DELEO J M. Artificial neural networks[J]. Cancer, 2001, 91(Suppl. 8): 1615-1635. http://d.old.wanfangdata.com.cn/Periodical/gpxygpfx201509047
    [21] SPECHT D F. A general regression neural network[J]. IEEE Transactions on Neural Networks, 1991, 2(6): 568-576. doi: 10.1109/72.97934
    [22] 田万银, 徐华潮.基于相空间重构及GRNN的海防林害虫预测及效果检验[J].浙江林业科技, 2014, 34(2): 65-69. doi: 10.3969/j.issn.1001-3776.2014.02.016

    TIAN W Y, XU H C. Coastal protection forest pest forecasting and verification based on phase space reconstruction and GRNN[J]. Journal of Zhejiang Forestry Science and Technology, 2014, 34(2): 65-69. doi: 10.3969/j.issn.1001-3776.2014.02.016
    [23] 金帅军.基于GRNN神经网络的农作物虫害量预测系统设计[D].呼和浩特: 内蒙古工业大学, 2013. http://cdmd.cnki.com.cn/Article/CDMD-10128-1014012754.htm

    JIN S J. Design the prediction of system insect pests of crop based on GRNN neural network[D]. Hohhot: Inner Mongolia University of Technology, 2013. http://cdmd.cnki.com.cn/Article/CDMD-10128-1014012754.htm
    [24] SVOZIL D, KVASNICKAB V, POSPICHALB J. Introduction to multi-layer feed-forward neural networks[J]. Chemometrics and Intelligent Laboratory Systems, 1997, 39(1): 43-62. doi: 10.1016/S0169-7439(97)00061-0
    [25] JOHANSSON E M, DOWLA F U, GOODMAN D M. Backpropagation learning for multilayer feed-forward neural networks using the conjugate gradient method[J]. International Journal of Neural Systems, 1991, 2(4): 291-301. doi: 10.1142/S0129065791000261
    [26] SPECHT D F. A general regression neural network[J]. IEEE Transactions on Neural Networks, 1991, 2(6): 568-576. doi: 10.1109/72.97934
    [27] GOULERMA J Y, LIATSIS P, ZENG X J, et al. Density-driven generalized regression neural networks (DD-GRNN) for function approximation[J]. IEEE Transactions on Neural Networks, 2007, 18(6): 1683-1696. doi: 10.1109/TNN.2007.902730
    [28] KHAN J, WEI J S, RINGNE M, et al. Classification and diagnostic prediction of cancers using gene expression profiling and artificial neural networks[J]. Nature Medicine, 2001, 7(6): 673-679. doi: 10.1038/89044
    [29] 琚存勇, 蔡体久.鄂尔多斯草地生物量估测的GRNN模型实现[J].北京林业大学学报, 2008, 30(增刊1): 296-299. http://j.bjfu.edu.cn/article/id/8516

    JU C Y, CAI T J. Estimation of aboveground biomass using GRNN model in Ordos grassland[J]. Journal of Beijing Forestry University, 2008, 30(Suppl. 1): 296-299. http://j.bjfu.edu.cn/article/id/8516
    [30] CORNE S A, CARVER S J, KUNIN W E, et al. Predicting forest attributes in southeast Alaska using artificial neural networks[J]. Forest Science, 2004, 50(2): 259-276.
    [31] KANDIEMAZ H M, KABA K, AVCI M. Estimation of monthly sunshine duration in turkey using artificial neural networks[J/OL]. International Journal of Photoenergy, 2014(2): 1-9[2016-05-10]. http://dx.doi.org/10.1155/2014/680596.
    [32] SCHOLKOPF B, SUNG K K, BURGES C, et al. Comparing support vector machines with Gaussian kernels to radial basis function classifiers[J]. IEEE Transactions on Signal Processing, 1997, 45(11): 2758-2765. doi: 10.1109/78.650102
    [33] DENG N Y, TIAN Y J, ZHANG C H. Support vector machines: optimization based theory, algorithms, and extensions[M]. 1st ed. New York: Chapman & Hall/CRC, 2012.
    [34] ZHONG X M, LI J P, DOU H C, et al. Fuzzy nonlinear proximal support vector machine for land extraction based on remote sensing image[J/OL]. PloS ONE, 2013, 8(7): e69434[2016-05-06]. DOI: 10.1371/journal.pone.0069434.
    [35] SHEN Y, HE Z, WANG Q, et al. Feature generation of hyperspectral images for fuzzy support vector machine classification[C]. Graz: IEEE Instrumentation and Measurement Technology Conference, 2012: 1977-1982.
    [36] 林卓, 吴承祯, 洪伟, 等.基于BP神经网络和支持向量机的杉木人工林收获模型研究[J].北京林业大学学报, 2015, 37(1): 42-47. doi: 10.13332/j.cnki.jbfu.2015.01.008

    LIN Z, WU C Z, HONG W, et al. Yield model of Cunninghamia lanceolata plantation based on back propagation neural network and support vector machine[J]. Journal of Beijing Forestry University, 2015, 37(1): 42-47. doi: 10.13332/j.cnki.jbfu.2015.01.008
    [37] YANG D, LI H, CHEN F, et al. A novel qualitative proof approach of the Dulong-Petit law using general regression neural networks[C]. Ottwa: 2014 IEEE Electronics, Computer and Applications, 2014: 577-580.
    [38] 罗瑜.支持向量机在机器学习中的应用研究[D].成都: 西南交通大学, 2007. http://cdmd.cnki.com.cn/article/cdmd-10613-2008041983.htm

    LUO Y. Study on application of machine learning based on support vector machine[D]. Chengdu: Southwest Jiaotong University, 2007. http://cdmd.cnki.com.cn/article/cdmd-10613-2008041983.htm
    [39] 费海泽, 王鸿斌, 孔祥波, 等.马尾松毛虫发生相关气象因子筛选及预测[J].东北林业大学学报, 2014, 42(1): 136-140. doi: 10.3969/j.issn.1000-5382.2014.01.030

    FEI H Z, WANG H B, KONG X B, et al. Selection and prediction of meteorological factors correlated with Dendrolimus punctatus outbreak[J]. Journal of Northeast Forestry University, 2014, 42(1): 136-140. doi: 10.3969/j.issn.1000-5382.2014.01.030
    [40] 陈梅香.基于SVM和GIS的梨小食心虫预测系统的研究[D].北京: 北京林业大学, 2010. http://cdmd.cnki.com.cn/Article/CDMD-10022-2010031389.htm

    CHEN M X. Studies on prediction system of Grapholitha molesta(Busck) based on support vector machine and geographic information system[D]. Beijing: Beijing Forestry University, 2010. http://cdmd.cnki.com.cn/Article/CDMD-10022-2010031389.htm
    [41] 向昌盛, 周子英, 张林峰.支持向量机在害虫发生量预测中的应用[J].生物信息学, 2011, 9(1): 28-31. doi: 10.3969/j.issn.1672-5565.2011.01.007

    XIANG C S, ZHOU Z Y, ZHANG L F. Application of pest occurrence prediction based on support vector machine[J]. China Journal of Bioinformatics, 2011, 9(1): 28-31. doi: 10.3969/j.issn.1672-5565.2011.01.007
  • 加载中
图(7) / 表(6)
计量
  • 文章访问数:  1263
  • HTML全文浏览量:  227
  • PDF下载量:  28
  • 被引次数: 0
出版历程
  • 收稿日期:  2016-06-20
  • 修回日期:  2016-10-23
  • 刊出日期:  2017-01-01

目录

    /

    返回文章
    返回