不同污染源海水样本的分类模型研究  

Study on Classification Model of Seawater Samples with Different Pollution Sources

作　　者：林燕[1] 朱尔一[1] 徐晓琴[1] LEE Frank S C 王小如[2] 

机构地区：[1]厦门大学化学化工学院,现代分析科学教育部重点实验室,福建厦门361005 [2]国家海洋局第一海洋研究所,山东青岛266061

出　　处：《光谱学与光谱分析》2007年第10期2107-2110,共4页Spectroscopy and Spectral Analysis

基　　金：福建省自然科学基金项目(Z0513003);海洋环境污染物被动示踪研究和国家"863"计划(2003AA635180)资助

摘　　要：简要介绍了偏最小二乘建立分类模型方法的原理及优点,将该方法应用于处理胶州湾和莱州湾的几个主要污染源附近海域各站点取得的海水样本的气-质联用全谱数据,建立海水样品的分类模型,以判别海水中有机污染物的来源区域。结果表明:由于PLS法适合于处理变量数多样本数少、具有严重多重共线性数据的问题,应用于从两类及多类海水样品气-质联用全谱数据中提取海水污染来源区域的分类信息,得到的分类模型交叉检验相关系数达0.91以上,结果较为理想,可为正确判别污染源提供一个可靠的基础。另外文章采用所得模型的拟合值等一些信息作分类图的方法,与传统PLS作图方法比较,所得分类图更为清晰、直观,能较好地表达回归模型的分类效果。In the present article the principle and advantages of the method to build classification model by partial least squares are briefly introduced. The method was applied to deal with the seawater data obtained from the primary polluted sea area of Jiaozhou bay and Laizhou bay by GC-MS. The classification models have been built for seawater samples from different contami- nated areas, The results indicate that PLS is very suitable for dealing with the problems with the data sets that contain many variables and few samples and have serious co-linearity. Accurate classification models can be built by use of PLS to get the classification information of pollution sources from two or many kinds of polluted seawaters data sets from GC-MS. The cross validation relativities of the model comes to over 0.91. This result is approving, which can provide a reliable foundation for distinguishing pollution sources correctly. Moreover, compared with the traditional method, the classification figures constructed by model＇s yi in the article are more clear and intuitive, and can express the model＇ s discrimination effect better.

关 键 词：气-质联用 偏最小二乘(PLS) 分类模型 

分 类 号：O657.6[理学—分析化学] O657.7[理学—化学]