Machine Learning to Predict the Adsorption Capacity of Microplastics.

Area of Science:

• Environmental Chemistry
• Polymer Science
• Computational Chemistry

Background:

• Extensive plastic production leads to microplastic and nanoplastic contamination in ecosystems.
• Microplastics in aquatic environments facilitate the adsorption and dispersal of chemical pollutants.

Purpose of the Study:

• To address the lack of data on microplastic adsorption of chemical pollutants.
• To develop predictive models for microplastic/water partition coefficients (log Kd).

Main Methods:

• Development of three machine learning models: random forest, support vector machine, and artificial neural network.
• Utilized two approximations based on the number of input variables for model training.
• Validated model performance using correlation coefficients in the query phase.

Main Results:

• The developed machine learning models achieved high predictive accuracy.
• Correlation coefficients above 0.92 were observed in the query phase for the best models.
• Demonstrated the capability of machine learning to estimate organic contaminant absorption on microplastics.

Conclusions:

• Machine learning models offer a rapid and reliable method for estimating organic contaminant absorption on microplastics.
• These models can significantly contribute to assessing the environmental risks associated with microplastic pollution.
• Further research can refine these models for broader applications in environmental monitoring.