United StatesPublic sector2015 – ongoingEnvironmental Protection Agency
Machine-learning targeting of hazardous-waste inspections
EPA reported nearly doubled violation detection with machine-learning inspection targeting
Problem
EPA had limited resources to inspect hazardous-waste facilities. It needed to identify facilities more likely to have serious violations.
Approach
EPA worked with the University of Chicago Energy & Environment Lab from 2015 to develop a predictive model using historical regulatory data. Regional field tests ran during fiscal years 2017 through 2019. Following a randomized evaluation, EPA released predictive scores to regions and states for inspection planning.
Results
Takeaway
The model supported inspectors' selection of facilities. EPA reported better detection without higher operating costs, rather than a measured reduction in pollution.
Categories
- Size
- Large enterprise
- Industry
- Public sector
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Partner co-development
- Evidence
- Company's own claim
Sources
Source summaries
EPA described its hazardous-waste inspection model and randomized evaluation. It reported a nearly doubled detection rate and released predictive scores for regional and state inspection planning.
Full details are in the original sources.