文献类型： 外文期刊

作者： Wang, Dingcheng ¹ ; Wang, Maohua ² ; Qiao, Xiaoiun ³ ;

作者机构： 1.Nanjing Univ Informat Sci Technol, Inst Comp & Software, Nanjing 210044, Peoples R China

2.China Agr Univ, Key Lab Modern Precis Agr Syst Integrat Res, Minist Educ, Beijing 100083, Peoples R China

3.Natl Engn Res Ctr Informat Technol Agr, Beijing 100083, Peoples R China

关键词： LSSVMR; Greenhouse; Online; Sparsity, Modeling

期刊名称：COMPUTERS AND ELECTRONICS IN AGRICULTURE （ 影响因子：5.565； 五年影响因子：5.494 ）

ISSN： 0168-1699

年卷期： 2009 年 66 卷 1 期

页码：

收录情况： SCI

摘要： The greenhouse environment is an uncertain nonlinear system which classical modeling methods cannot solve. Support vector machines regression (SVMR) is well supported by mathematical theory and has a simple structure, good generalization ability, and nonlinear modeling properties. Therefore, SVMR offers a very competent method for modeling the greenhouse environment. However, to deal with uncertainty, the model must be rectified online, and Online Sparse Least-Squares Support Vector Machines Regression (OS_LSSVMR) was developed to solve this problem. OS_LSSVMR reduced the number of training samples through use of a sample dictionary, and consequently LSSVMR has sparse solutions; the training samples were added sequentially, so that OS_LSSVMR has online learning capability. A simplified greenhouse model, in which only greenhouse internal and external air temperatures were considered, was presented, after analyzing the factors in the greenhouse environment. Then the OS_LSSVMR greenhouse model was constructed using real-world data. The resulting model shows a promising performance in the greenhouse environment, with potential improvements if a more complete data setup is used. (C) 2008 Elsevier B.V. All rights reserved.