基于PSO-OSVM的江苏省耕地变化预测  

Forecast of the Cultivated Land Change of Jiangsu Province Based on PSO-OSVM Model

作　　者：何厅厅[1] 赵艳玲[1] 石娟娟[1] 刘亚萍[1] 王亚云[1] 袁军[1] 马和平[1] 

机构地区：[1]中国矿业大学(北京)土地复垦与生态重建研究所,北京100083

出　　处：《湖北农业科学》2014年第21期5315-5319,共5页Hubei Agricultural Sciences

基　　金：教育部新世纪优秀人才支持计划项目(NCET-12-0964);中央高校基本科研业务费专项资金资助项目(2009QD05)

摘　　要：以江苏省1996-2009年耕地变化为例,利用粒子群算法(PSO)的全局搜索能力优化标准支持向量机(SVM),并结合增量式最小二乘支持向量机(LSSVR)和逆学习算法的特征,构建粒子群算法-在线学习SVM(PSO-OSVM)耕地变化预测模型,采用该模型对江苏省耕地变化进行预测,以期为土地资源可持续发展提供重要参考依据。结果表明,PSO可以有效收敛SVM内部参数γ和σ达到全局最优解;PSOOSVM模型的内外精度和总精度均高于GM(1,1)、BP神经网络模型,且优于PSO-SVM模型。说明PSOOSVM是一种有效的耕地变化预测模型。With the change in cultivated land of Jiangsu province from 1996 to 2009 as an example with the optimized standard support vector machine （SVM） by using the global search ability of particle swarm optimization （PSO） combined with the characteristics of least squares support vector machine （LSSVM） with converse learning algorithm,PSO-OSVM forecast model of change in cultivated land was established and then adopted to predict change in cultivated land of Jiangsu province in order to provide an important reference for sustainable development.The results showed that PSO could effectively converge SVM internal parameters γ and σ to achieve the global optimal solution.The internal and external precision and total precision of PSO-OSVM were higher than GM （1,1） and BP neural network,and it was better than PSO-SVM.Result showed that the PSO-OSVM was an effective forecast model of cultivated land change.

关 键 词：在线学习支持向量机 粒子群算法 耕地变化预测 

分 类 号：S159[农业科学—土壤学] P209[农业科学—农业基础科学]