Assessment of machine learning-based methods predictive suitability for migration pollutants from microplastics...

Area of Science:

• Environmental Chemistry
• Computational Science

Background:

• Microplastics pose environmental risks due to pollutant leaching.
• Reducing laboratory analyses is crucial for economic and ecological reasons.

Purpose of the Study:

• To evaluate machine learning for predicting microplastic pollutant migration.
• To explore cost-effective and environmentally conscious analytical methods.

Main Methods:

• Utilized multiple regression, artificial neural networks, support vector method, and random forest regression.
• Developed models based on gas chromatography-mass spectrometry (GC-MS) laboratory data.
• Focused on predicting the leaching of plasticizers and other contaminants.

Main Results:

• Artificial neural networks and support vector methods showed high predictive accuracy (correlations 0.96-0.99).
• Multiple regression demonstrated low performance in predicting phthalic acid esters (R² 0.03-0.24).
• This study is the first to apply machine learning to predict plasticizer leaching from various polymers.

Conclusions:

• Artificial neural networks and support vector methods are effective for modeling chemical leaching from microplastics.
• Machine learning offers a viable alternative to extensive laboratory testing.
• Findings provide critical data for assessing the environmental impact of microplastic additives.