Forecasting toxic metal concentrations in an inland sea ecosystem with machine learning algorithms

Area of Science:

• Environmental Science
• Data Science
• Geochemistry

Background:

• Statistical and data-driven modeling are crucial for environmental feature analysis.
• Feature selection integrated with machine learning improves model generalization and reduces complexity.
• Understanding element relationships is key to predicting environmental concentrations.

Purpose of the Study:

• To investigate relationships between Aluminum (Al) and other elements.
• To predict Al concentration levels in an inland marine ecosystem using machine learning.
• To evaluate the efficacy of using a reduced feature set versus a full set for accurate Al prediction.

Main Methods:

• Application of machine learning models for predictive analysis.
• Integration of feature selection techniques to identify significant elements.
• Comparative analysis of model performance using full and reduced feature sets.

Main Results:

• Machine learning models successfully predicted Al concentrations.
• Feature selection identified a subset of informative elements crucial for prediction.
• Models utilizing reduced feature sets yielded accurate and more interpretable results.

Conclusions:

• Machine learning combined with feature selection is effective for predicting Al concentrations.
• A reduced set of significant elements enhances model interpretability and predictive power.
• This approach offers a robust method for environmental element analysis in marine ecosystems.