[Study on classification model of seawater samples with different pollution sources].

Area of Science:

• Environmental Chemistry
• Chemometrics
• Analytical Chemistry

Context:

• Seawater quality assessment in coastal areas like Jiaozhou and Laizhou bays.
• Challenges in analyzing complex datasets from Gas Chromatography-Mass Spectrometry (GC-MS).
• Need for robust methods to identify multiple pollution sources.

Purpose:

• Introduce the principle and advantages of Partial Least Squares (PLS) for classification modeling.
• Apply PLS to GC-MS data for classifying seawater samples from polluted marine environments.
• Evaluate the effectiveness of PLS in handling datasets with high dimensionality and collinearity.

Summary:

• Partial Least Squares (PLS) models were developed using GC-MS data from polluted seawater samples.
• The PLS method demonstrated suitability for datasets with many variables and few samples, and serious co-linearity.
• Accurate classification models were achieved, with cross-validation reliabilities exceeding 0.91, enabling reliable pollution source identification.

Impact:

• Provides a reliable foundation for distinguishing marine pollution sources.
• Offers a more clear, intuitive, and effective method for visualizing classification results compared to traditional approaches.
• Highlights the utility of PLS in environmental data analysis and pollution monitoring.