Machine Learning Pattern Recognition for Forensic Analysis of Detected Per- and Polyfluoroalkyl Substances in Environmental Samples
Abstract
This proof of concept project explored the use of modern machine learning algorithms for forensic source allocation of detected per- and polyfluoroalkyl substances (PFAS) in environmental samples. A challenge with allocating PFAS source by composition is that PFAS composition varies substantially even from a single source, as a result of differing compound mobility, as well as transformation of precursor compounds. Work involved compiling an extensive dataset from a range of worldwide PFAS concentration data sources, and then using a portion of the dataset to train supervised machine learning classifiers to recognize patterns that can distinguish between PFAS from AFFF and non-AFFF sources. Results show that a number of very different supervised learning classifiers work extremely well for this purpose, and are able to identify AFFF in samples even in difficult-to-recognize categories of samples. The results of this project show that supervised machine learning has significant promise as a means of distinguishing between PFAS from AFFF and non-AFFF sources.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 2020
- Accession Number
- AD1154374
Entities
People
- Denis M O'Carroll
- Rafal Jabrzemski
- Tohren Kibbey
Organizations
- University of Oklahoma