应用随机森林模式识别的土壤肥力评价
|
摘要
针对目前以线性函数描述指标因子评价土壤肥力量级时,因权重分配会受到数据噪声影响,以及在肥力量级划分的过程中存在较大主观性的问题,基于机器学习模式构建了一种土壤肥力的自动化评测系统.应用随机森林模式识别理论,通过深入挖掘土壤养分与土壤肥力之间的非线性映射关系获取学习能力,以分类树为基元,将土壤养分指标以叶节点输入,以肥力量级从根节点输出,建立了土壤肥力质量评测模型.然后基于现有的土壤质量分级标准,生成3 000组训练样本对模型进行训练,最后将训练好的模型对试区64个土壤样点进行预测分析.实验结果表明,试区土壤化学肥力量级分属Ⅱ至Ⅴ级,以Ⅲ和Ⅳ级分布占优,与实际情况相符,证明所建立的土壤肥力自动评测系统能够实现区域土壤的自动化评测,且分类精度高、过程客观.
In view of the evaluation of soil fertility level,the current using linear function with index factor method is affected by data noise and the subjective problem in the process of fertility levels division.An automated evaluation system of soil fertility level was constructed based on machine learning mode.The system applied the theory of Random forest pattern recognition,taking the classification tree as the base element,the soil nutrient index as the leaf node,and the soil fertility level as the output from the root node.The learning ability of this system was acquired by deepening the nonlinear mapping relationship between soil nutrient and soil fertility level.Based on the existing soil quality grading standards,3 000 sets of training samples were generated to train the system.Finally,the trained model was used to predict and analyze 64 soil samples in the test area.The experimental results show that the soil fertility level in the test area belongs to II to V,which are dominant in the III and IV distributions.It is consistent with the actual situation.The automatic soil fertility evaluation system established in this paper can realize the automatic evaluation of regional soil.The classification accuracy is high and the process is objective.
In view of the evaluation of soil fertility level,the current using linear function with index factor method is affected by data noise and the subjective problem in the process of fertility levels division.An automated evaluation system of soil fertility level was constructed based on machine learning mode.The system applied the theory of Random forest pattern recognition,taking the classification tree as the base element,the soil nutrient index as the leaf node,and the soil fertility level as the output from the root node.The learning ability of this system was acquired by deepening the nonlinear mapping relationship between soil nutrient and soil fertility level.Based on the existing soil quality grading standards,3 000 sets of training samples were generated to train the system.Finally,the trained model was used to predict and analyze 64 soil samples in the test area.The experimental results show that the soil fertility level in the test area belongs to II to V,which are dominant in the III and IV distributions.It is consistent with the actual situation.The automatic soil fertility evaluation system established in this paper can realize the automatic evaluation of regional soil.The classification accuracy is high and the process is objective.
引文
[1]Cardoso I M,Kuyper T W.Mycorrhizas and tropical soil fertility[J].Agriculture Ecosystems &Environment,2006,116(1):72-84.
[2]骆东奇,白洁,谢德体.论土壤肥力评价指标和方法[J].生态环境学报,2002,11(2):202-205.Luo Dongqi,Bai Jie,Xie Deti.Research on evaluation norm and method of soil fertility[J].Soil and Environmental Sciences,2002,11(2):202-205.(in Chinese)
[3]刘洪鹄,赵玉明,王秀颖,等.土壤肥力评价方法探讨[J].长江科学院院报,2008,25(3):62-66.Liu Honghu,Zhao Yuming,Wang Xiuying,et al.Diseussion of evaluation methods on soil fertility[J].Journal of Yangtze River Scientific Research Institute,2008,25(3):62-66.(in Chinese)
[4]庞元明.土壤肥力评价研究进展[J].山西农业科学,2009,37(2):85-87.Pang Yuanming.Progress in soil fertility evaluation[J].Shanxi Agricultural Sciences,2009,37(2):85-87.(in Chinese)
[5]颜雄,张杨珠,刘晶.土壤肥力质量评价的研究进展[J].湖南农业科学,2008(5):82-85.Yan Xiong,Zhang Yangzhu,Liu Jing.Progress of evaluation methods on soil fertility[J].Hunan Agricultural Sciences,2008(5):82-85.(in Chinese)
[6]李润林,姚艳敏,董鹏程.基于AHP、负权重和模糊数学的土壤质量评价研究[J].湖北农业科学,2016,55(17):4480-4483.Li Runlin,Yao Yanmin,Dong Pengcheng.The assessment of soil quality based on AHP,negative weights and fuzzy math[J]. Hubei Agricultural Sciences,2016,55(17):4480-4483.(in Chinese)
[7]姜洪伟.黑土肥力的演变过程和评价指标体系研究[J].现代农业科技,2011(6):278-279.Jiang Hongwei.Research on evolution processes and evaluation index system of black son fertility[J].Modern Agricultural Technology,2011(6):278-279.(in Chinese)
[8]冉延平.城市表层土壤重金属污染多元统计分析和污染评价[J].安徽农业科学,2012,40(9):5454-5458.Ran Yanping.Multivarlate statistical analysis of heavy metats contamination in urinm topsoil and tinn assessment[J].Journal of Anhui Agricultural Sciences,2012,40(9):5454-5458.(in Chinese)
[9]何平平,康可佳,宋利臣.基于主成分分析的大兴安岭高寒区典型火烧迹地土壤质量评价研究[J].环境科学与管理,2015,40(6):170-173.He Pingping,Kang Kejia,Song Lichen.Evaluation of burned soil quality in typical alpine area of greater Xi’an mountains based on principal component analysis[J].Environmental Science and Management,2015,40(6):170-173.(in Chinese)
[10]彭景,李泽琴,侯家渝.地积累指数法及生态危害指数评价法在土壤重金属污染中的应用及探讨[J].广东微量元素科学,2007,14(8):13-17.Peng Jing,Li Zeqin,Hou Jiayu.Appfication of the index of geo-accumulation index and ecological risk in dex to assess heavy metal pollution in soils[J].Guangdong Trace Elements Science,2007,14(8):13-17.(in Chinese)
[11]杨芳,张江山,王菲凤.基于熵权的属性识别模型在土壤重金属污染评价中的应用[J].环境保护科学,2008,34(6):57-60.Yang Fang,Zhang Jiangshan,Wang Feifeng.Application of attribute recognition model based on coefficient of entropy to comprehensive evaluation of soil heavy metal pollution[J].Environmental Protection Science,2008,34(6):57-60.(in Chinese)
[12]陈伯扬.TOPSIS法在土壤环境质量评价中的应用[J].现代地质,2008,22(6):1003-1009.Chen Boyang.Application of TOPSIS method to the environment quality evaluation of soil[J].Geoscience,2008,22(6):1003-1009.(in Chinese)
[13]陈凯,朱钰.机器学习及其相关算法综述[J].统计与信息论坛,2007,22(5):105-112.Chen Kai,Zhu Yu.A summary of macllme learning and related algorithms[J].Statistics &Information Forum,2007,22(5):105-112.(in Chinese)
[14]南京农学院.土壤农化分析[M].北京:农业出版社,1980.
[15]Breiman L.Random forests[J].Machine Learning,2001,45(1):5-32.
[16]Cutler D R,Edwards T C,Beard K H,et al.Random forests for classification in ecology[J].Ecology,2007,88(11):2783-2792.
[17]GB 15618-1995土壤环境质量标准[S].北京:中国环境科学出版社.
[18]Cutler A,Cutler D R,Stevens J R.Random forests[J].Machine Learning,2004,45(1):157-176.
[19]赵北庚.基于R语言randomForest包的随机森林建模研究[J].计算机光盘软件与应用,2015,18(2):152-153.Zhao Beigeng.Research on random forest modeling based on R language random forest package[J].Computer CD Software and Applications,2015,18(2):152-153.(in Chinese)
[20]张颖,高倩倩.基于随机森林分类算法的巢湖水质评价[J].环境工程学报,2016,10(2):992-998.Zhang Ying,Gao Qianqian.Water quality evaluation of Chaohu Lake based on random forest classification algorithm[J].Journal of Environmental Engineering,2016,10(2):992-998.(in Chinese)
[21]Chen C,Classification I.Topics in random forests[D]. Berkeley:University of California at Berkeley,2005.
[2]骆东奇,白洁,谢德体.论土壤肥力评价指标和方法[J].生态环境学报,2002,11(2):202-205.Luo Dongqi,Bai Jie,Xie Deti.Research on evaluation norm and method of soil fertility[J].Soil and Environmental Sciences,2002,11(2):202-205.(in Chinese)
[3]刘洪鹄,赵玉明,王秀颖,等.土壤肥力评价方法探讨[J].长江科学院院报,2008,25(3):62-66.Liu Honghu,Zhao Yuming,Wang Xiuying,et al.Diseussion of evaluation methods on soil fertility[J].Journal of Yangtze River Scientific Research Institute,2008,25(3):62-66.(in Chinese)
[4]庞元明.土壤肥力评价研究进展[J].山西农业科学,2009,37(2):85-87.Pang Yuanming.Progress in soil fertility evaluation[J].Shanxi Agricultural Sciences,2009,37(2):85-87.(in Chinese)
[5]颜雄,张杨珠,刘晶.土壤肥力质量评价的研究进展[J].湖南农业科学,2008(5):82-85.Yan Xiong,Zhang Yangzhu,Liu Jing.Progress of evaluation methods on soil fertility[J].Hunan Agricultural Sciences,2008(5):82-85.(in Chinese)
[6]李润林,姚艳敏,董鹏程.基于AHP、负权重和模糊数学的土壤质量评价研究[J].湖北农业科学,2016,55(17):4480-4483.Li Runlin,Yao Yanmin,Dong Pengcheng.The assessment of soil quality based on AHP,negative weights and fuzzy math[J]. Hubei Agricultural Sciences,2016,55(17):4480-4483.(in Chinese)
[7]姜洪伟.黑土肥力的演变过程和评价指标体系研究[J].现代农业科技,2011(6):278-279.Jiang Hongwei.Research on evolution processes and evaluation index system of black son fertility[J].Modern Agricultural Technology,2011(6):278-279.(in Chinese)
[8]冉延平.城市表层土壤重金属污染多元统计分析和污染评价[J].安徽农业科学,2012,40(9):5454-5458.Ran Yanping.Multivarlate statistical analysis of heavy metats contamination in urinm topsoil and tinn assessment[J].Journal of Anhui Agricultural Sciences,2012,40(9):5454-5458.(in Chinese)
[9]何平平,康可佳,宋利臣.基于主成分分析的大兴安岭高寒区典型火烧迹地土壤质量评价研究[J].环境科学与管理,2015,40(6):170-173.He Pingping,Kang Kejia,Song Lichen.Evaluation of burned soil quality in typical alpine area of greater Xi’an mountains based on principal component analysis[J].Environmental Science and Management,2015,40(6):170-173.(in Chinese)
[10]彭景,李泽琴,侯家渝.地积累指数法及生态危害指数评价法在土壤重金属污染中的应用及探讨[J].广东微量元素科学,2007,14(8):13-17.Peng Jing,Li Zeqin,Hou Jiayu.Appfication of the index of geo-accumulation index and ecological risk in dex to assess heavy metal pollution in soils[J].Guangdong Trace Elements Science,2007,14(8):13-17.(in Chinese)
[11]杨芳,张江山,王菲凤.基于熵权的属性识别模型在土壤重金属污染评价中的应用[J].环境保护科学,2008,34(6):57-60.Yang Fang,Zhang Jiangshan,Wang Feifeng.Application of attribute recognition model based on coefficient of entropy to comprehensive evaluation of soil heavy metal pollution[J].Environmental Protection Science,2008,34(6):57-60.(in Chinese)
[12]陈伯扬.TOPSIS法在土壤环境质量评价中的应用[J].现代地质,2008,22(6):1003-1009.Chen Boyang.Application of TOPSIS method to the environment quality evaluation of soil[J].Geoscience,2008,22(6):1003-1009.(in Chinese)
[13]陈凯,朱钰.机器学习及其相关算法综述[J].统计与信息论坛,2007,22(5):105-112.Chen Kai,Zhu Yu.A summary of macllme learning and related algorithms[J].Statistics &Information Forum,2007,22(5):105-112.(in Chinese)
[14]南京农学院.土壤农化分析[M].北京:农业出版社,1980.
[15]Breiman L.Random forests[J].Machine Learning,2001,45(1):5-32.
[16]Cutler D R,Edwards T C,Beard K H,et al.Random forests for classification in ecology[J].Ecology,2007,88(11):2783-2792.
[17]GB 15618-1995土壤环境质量标准[S].北京:中国环境科学出版社.
[18]Cutler A,Cutler D R,Stevens J R.Random forests[J].Machine Learning,2004,45(1):157-176.
[19]赵北庚.基于R语言randomForest包的随机森林建模研究[J].计算机光盘软件与应用,2015,18(2):152-153.Zhao Beigeng.Research on random forest modeling based on R language random forest package[J].Computer CD Software and Applications,2015,18(2):152-153.(in Chinese)
[20]张颖,高倩倩.基于随机森林分类算法的巢湖水质评价[J].环境工程学报,2016,10(2):992-998.Zhang Ying,Gao Qianqian.Water quality evaluation of Chaohu Lake based on random forest classification algorithm[J].Journal of Environmental Engineering,2016,10(2):992-998.(in Chinese)
[21]Chen C,Classification I.Topics in random forests[D]. Berkeley:University of California at Berkeley,2005.