The problem
Exposure averages can obscure peaks and episodes. This project represents local time-series structure so it can be investigated alongside conventional summaries.
My contribution
I worked on time-series preparation, pattern extraction, and reusable representations within the SMARTER project. Our team developed a web system to explore and retrieve these patterns.
The approach
- Organize EPA monitor data into pollutant time series.
- Extract 7- and 30-day candidate subsequences.
- Use exposure-aware criteria to retain high-quality, non-redundant patterns.
- Support pattern retrieval and downstream feature construction.
What this demonstrates
Method development, scientific data engineering, and translating a research representation into a reusable resource.
Interpretation: This is a representation and retrieval framework. Clinical predictive benefit must be evaluated separately for each downstream task.
The research system is hosted by the University of Utah; availability depends on the host.