Study on classification model of Seawater samples with different pollution sources
Lee Frank S C
- 化学化工－已发表论文 
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 contaminated 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 y(i) in the article are more clear and intuitive, and can express the model's discrimination effect better.