详细信息
粒子群支持向量机结合NIR测定桉木木质素 被引量:5
PSO Support Vector Machine Combined with NIR Determination for the Lignin Content of Eucalyptus
文献类型:期刊文献
中文题名:粒子群支持向量机结合NIR测定桉木木质素
英文题名:PSO Support Vector Machine Combined with NIR Determination for the Lignin Content of Eucalyptus
作者:于仕兴[1] 李学春[2] 黄安民[3] 王学顺[1]
第一作者:于仕兴
机构:[1]北京林业大学;[2]广东省水产学校;[3]中国林业科学研究院木材工业研究所
年份:2013
卷号:41
期号:2
起止页码:123-126
中文期刊名:东北林业大学学报
外文期刊名:Journal of Northeast Forestry University
收录:CSTPCD;;北大核心:【北大核心2011】;CSCD:【CSCD2013_2014】;
基金:北京市自然科学基金项目(6092021);“十一五”国家科技支撑计划项目(2006BAD03A15)
语种:中文
中文关键词:支持向量机;近红外光谱;粒子群优化算法;木质素;回归
外文关键词:Support vector machines (SVM) ; Near-infrared spectroscopy; Particle swarm optimization (PSO) ; Lignin ; Regression
分类号:O29
摘要:在支持向量机(SVM)回归分析过程中,参数(C,γ)取值范围较大,且需要人工进行调整,目前已知的参数选择方法复杂且不够精确。针对上述问题,提出了一种应用于木材近红外光谱分析的PSO-SVM回归模型;使用粒子群算法(PSO)确定SVM的最优参数(C,γ),用40个桉木近红外光谱样品作训练集,8个样品作测试集建立模型,得到预测模型的回归系数0.970 956,均方根误差0.002 154 5,并与传统支持向量机回归模型和偏最小二乘回归模型进行分析比较。结果表明,PSO-SVM回归模型在桉木近红外光谱的木质素含量预测中具有较高的准确性和很好的稳定性。
In the analysis process of support vector machine (SVM) regression, the parameters ( G,y) are m the larger range and need manual adjustment. However, the parameter selection method is complex and imprecise now. The SVM-PSO ( particle swarm optimization) regression model was advanced to determine the optimal parameters (C,y) with 40 samples of eucalyptus near-infrared spectra for the training set and 8 samples for testing set model. The regression coefficient of prediction model was 0.970 956, and root mean square error was 0.002 154 5. The model was compared with the traditional support vector machine regression model and partial least squares regression analysis. The results showed that PSO-SVM regression model is with the advantages of high accuracy and good stability in forecast lignin content of the eucalyptus nearinfrared spectrum.
参考文献:
正在载入数据...