Machine learning: Next promising trend for microplastics study.

Area of Science:

• Environmental Science
• Data Science
• Analytical Chemistry

Background:

• Microplastics (MPs) are pervasive pollutants with significant ecological and health risks.
• Traditional MP characterization methods are time-consuming and require substantial sample volumes.
• Machine Learning (ML) presents a powerful alternative for MP analysis due to its accuracy and feature extraction capabilities.

Purpose of the Study:

• To provide a comprehensive overview of ML applications in microplastic research.
• To address the knowledge gap and imbalanced development of ML in the environmental science community.
• To guide researchers in dataset construction, algorithm selection, and performance evaluation for MP studies.

Main Methods:

• Literature review focusing on MPs datasets, ML algorithms, and characterization techniques.
• Categorization of ML-based MP identification into spectral, image, and spectral imaging methods.
• Discussion of ML applications beyond identification, including toxicity, adsorption, and microbial studies.

Main Results:

• Analysis of MPs datasets and common ML algorithms regarding interpretability and computational needs.
• Exploration of methods for enhancing ML model performance, including pre-processing and optimization.
• Identification of three primary categories for ML-based MP identification: spectral, image, and spectral imaging.

Conclusions:

• ML holds significant potential for advancing microplastic research across various applications.
• A proposed algorithm selection strategy aims to improve efficiency and reduce trial-and-error costs for researchers.
• This review facilitates broader and more effective implementation of ML in microplastic studies.