期刊文献+

不同数据挖掘方法在综合干旱监测模型构建中的应用研究 被引量:12

A Study of Application of Different Data Mining methods in Integrated Drought Monitoring
下载PDF
导出
摘要 为探讨数据挖掘方法和建模因变量对区域干旱综合监测模型构建的影响,选择分类回归树(CART)、人工神经网络(ANN)和支持向量机(SVM)等不种数据挖掘方法,以及以标准化降水指数(SPI)、标准化降水蒸散指数(SPEI)和综合气象干旱指数(CI)等不同指数为建模因变量,以山西省为研究区域,开展综合干旱监测模型研究,结果表明,所有构建模型的监测结果与模型因变量计算值之间相关系数在0.800~0.915,均达到极显著水平,均方根误差(RMSE)在0.061~0.138,并与土壤相对湿度的相关性均达到显著和极显著水平,但不同数据挖掘方法和不同干旱指数作为因变量构建的模型仍存在差异,且模型参数的选取对于模型泛化能力有较大的影响;从山西省干旱监测结果分析,SVM方法建立的模型表现出对高程较高的敏感性,同时作为模型因变量的干旱指数的特征也会反映在模型的应用效果中,在干旱等级划分指标标准相同的情况下,不同的数据挖掘方法和建模因变量构建的模型,监测到的干旱面积存在一定差异。 To study the influence of data mining methods and the dependent variable for modeling on regional drought monitoring model building,integrated drought monitoring models were studied taking Shanxi Province as the study area.Different data mining methods of classification and regression tree( CART),support vector machines( SVM) and artificial neural networks( ANN) were selected and the standardized precipitation evapotranspiration index( SPEI) and CI were selected as the dependent variables respectively for modeling.The results showed that correlation coefficient between the results obtained by all the models and the the actual calculated values of dependent variables was 0. 800-0. 915,all above the extremely significant level,and the RMSE was between 0. 061 and 0. 138. The correlation coefficient between the results obtained by each model and the soil moisture measured by the agricultural stations all reached a significant or extremely significant level.But there were some differences in the models developed by different data mining methods and different drought indices as the dependent variable for modeling. Furthermore,the parameters had an important influence on the generalization ability of the models.Then the results of drought monitoring in Shanxi Province showed that these models established with SVM are highly sensitive to the elevation.The characteristics of the drought index as dependent variable are also reflected in the application of the models. There are differences in the monitoring drought area under the same drought index classification with different data mining methods and dependent variables while constructing models.
出处 《江西农业大学学报》 CAS CSCD 北大核心 2017年第5期1047-1056,共10页 Acta Agriculturae Universitatis Jiangxiensis
基金 国家自然科学基金重点支持项目(91437220)~~
关键词 数据挖掘 干旱监测 综合干旱指数 data mining drought monitoring drought indices
  • 相关文献

参考文献7

二级参考文献125

共引文献186

同被引文献205

引证文献12

二级引证文献100

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部