摘要
以云南省曲靖市最严格水资源管理评价为研究对象,提出了最严格水资源管理评价指标体系和分级标准,构建基于回归支持向量机(SVR)和径向基函数(RBF)神经网络的评价模型。利用层次分析法(AHP)从用水总量、用水效率、限制纳污与责任考核4个方面遴选出20个指标,构建最严格水资源管理评价指标体系和分级标准;采用随机生成和随机选取的方法在最严格水资源管理评价等级标准阈值间构造小容量训练样本和检验样本对SVR与RBF模型进行验证。利用SVR与RBF模型对实例进行评价分析。结果表明:1SVR与RBF模型具有较高的评价精度和泛化能力,可用于最严格水资源管理评价。2SVR与RBF模型对曲靖市2010、2015、2020和2030年最严格水资源管理评价分别为"不理想","较理想","理想"和"最理想"。
Taking the most strict water resources management in Qujing of Yunnan province as an exam-ple ,the paper put forward evaluation index system and grading standard of the most strict water resources management and constructed the model of evaluation based on support vector machine regression ( SVR) and radial basis function ( RBF) neural network .By use of analytic hierarchy process ( AHP) ,it selected 20 indicators from 4 aspects such as total water use , water use efficiency , limiting pollutant and responsi-bility assessment , constructed the evaluation index system and grading standards of the most strict water resources management;and used the method of randomly generating and randomly selecting to verify the SVR model and RBF model in the most strict water resources management evaluation grade standard threshold between the construction of small capacity of training and testing samples .It evaluated and ana-lyzed the example by use of SVR and RBF model .The results showed that SVR and RBF models have higher evaluation accuracy and generalization ability , and can be used for the most strict water resources management and evaluation .The evaluation results by SVR and RBF model for the most strict water re-sources management of Qujing in 2010 , 2015 , 2020 and 2030 are “not ideal”,“more ideal”,“ideal”and “the most ideal”.
出处
《水资源与水工程学报》
2015年第2期119-125,共7页
Journal of Water Resources and Water Engineering
关键词
水资源
最严格水资源管理
指标体系
回归支持向量机
径向基函数神经网络
曲靖市
water resources
the most stringent water management
indicator system
regression support vector machine
radial basis function neural network
Qujing
