动态优化偏最小二乘模型的建立与应用  

Dynamic Optimized Partial Least Squares Regression

作　　者：杨岚[1] 冯新泸[1] 

机构地区：[1]后勤工程学院军事油料应用与管理工程系,重庆400016

出　　处：《后勤工程学院学报》2008年第2期75-77,共3页Journal of Logistical Engineering University

摘　　要：介绍了动态优化方法选择偏最小二乘法最佳主成分数的原理，以汽油烯烃指标为检测对象，研究了主成分数对检测结果的影响。研究结果表明：应用交互验证方法推荐的主成分数并不是最优，通过动态优化方法确定主成分数可提高所建立数学模型的预测效果，与交互验证方法选择主成分数方法相比较，动态优化方法确定的主成分数能够得到更好的预测结果。该方法能有效提高偏最小二乘数学模型的预测效果，是建立具有更好适应性数学模型的有效方法。This paper introduces the principle of dynamic optimized Partial Least Squares Regression. The effect of the principal component number on results has been studied by using the gasoline olefin as the detecting objects. Study results show that using the principal component number of Gasoline olefin model recommended by the method of cross-validation is not optimal and using the method of dynamic optimization to identify principal component number can improve predicted results. The dynamic optimization to identify principal component number can obtain better predicted results than cross-validation. This method can not only effectively improve the predicted effect of the PLS mathematical model, but also is an effective way to setting up a more adaptable mathematical model.

关 键 词：主成分数 动态优化 匹配 偏最小二乘 

分 类 号：O433.4[机械工程—光学工程]