Machine Learning Classification of PFAS

Abstract Submission: Per- and polyfluoroalkyl substances (PFAS) are a group of manufactured chemicals that pose serious environmental health risks due to their persistence and widespread occurrence in ecosystems. They can accumulate in soil, water, and organisms, leading to significant health concerns from long-term exposure. PFAS contamination originates from various sources, including firefighting training facilities and wastewater treatment plants, as well as from recycled consumer products like paper sludges used in agriculture. Identifying these sources is essential for effective remediation and environmental management, but the complex nature and varied origins of PFAS makes this challenging. Forensic methods are essential for identifying sources of PFAS contamination to better understand their behavior in the environment and mitigate exposure risks. Identifying these sources helps prevent future environmental releases, increases public awareness of exposure dangers, and supports the development of effective and sustainable remediation strategies. This paper explores the use of machine learning techniques to determine the sources of PFAS contamination by analyzing patterns in component concentrations from more than 300,000 different PFAS samples. Impacts of different preprocessing approaches and feature engineering on classification performance are evaluated. The use of machine learning could prove beneficial for forensic identification of PFAS sources by analyzing water sample compositions, aiding in source allocation strategies.

Learning Objectives/Expected Outcome (Optional) : Classification of PFAS, Machine Learning.