(51b) Applying Machine Learning to Estimate Releases from New Uses of Existing Chemicals | AIChE

(51b) Applying Machine Learning to Estimate Releases from New Uses of Existing Chemicals

Authors 

Smith, R. - Presenter, US Environmental Protection Agency
Hernandez-Betancur, J. D., U.S. Environmental Protection Agency
Meyer, D. E., U.S. Environmental Protection Agency
Ruiz-Mercado, G., U.S. Environmental Protection Agency
Barrett, W., US Environmental Protection Agency
Gonzalez, M., U.S. Environmental Protection Agency
Abraham, J. P., U.S. Environmental Protection Agency
New uses of existing chemicals may present unanticipated health risks. These risks have to be evaluated according to the Toxic Substances Control Act as amended by the 2016 Frank R. Lautenberg Chemical Safety for the 21st Century Act. Such an evaluation can follow the human health risk assessment paradigm of chemical release, exposure, and hazard assessment. In this work developed at the U.S. Environmental Protection Agency’s Office of Research and Development, the emphasis is on estimating the release of a chemical employed in a new manner or use. Generic scenarios are models which relate a chemical’s properties, its usage, and an activity to that chemical’s release. When information about the use of a chemical of interest in such activities is new, and therefore not available, one approach to consider is applying machine learning to estimate the releases. This work applies machine learning from both the chemical and use perspectives to estimate releases for new uses of existing chemicals. Efforts to access and preprocess data, develop and train the model, and test the model will be described, along with prediction examples for chemicals and new uses.

The views expressed in this presentation are those of the authors and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency.