Operational Data Analysis
Time-Aware Tables and Windows
Resample series, align calendars, and compute rolling metrics without silent gaps.
Tuition is informational only. Enrollment staff confirm availability and policies by email.
Request a syllabus callOverview
Operational metrics rarely arrive on neat grids. You will work with business-day offsets, late-arriving rows, and windowed aggregates that stay stable when clocks change. Emphasis is on readable code reviewers can audit.
What you build
- DatetimeIndex alignment patterns
- Rolling and expanding transforms with explicit windows
- Handling duplicate timestamps without silent drops
- Visualization of gaps using small multiples
- Unit tests for edge cases around midnight boundaries
- Export to Parquet with preserved time zones
- Documentation snippets for downstream dashboards
Outcomes
- Reproduce a month-end close metric with documented assumptions
- Flag two edge cases where naive rolling math would mislead readers
- Publish a short design note alongside your notebook
Mentor on record
Haneul Park
Lead Python instructor focused on analyst-friendly tooling and reviewable notebooks.
FAQ
Which libraries are emphasized?
Primarily pandas with selective NumPy. We mention Polars only where it clarifies performance trade-offs, not as a full migration course.
Is homework graded?
You receive structured comments on two milestones. Remaining exercises are self-paced with reference solutions.
Limitations?
We do not cover stream processing frameworks. Batch workflows are the scope boundary.
Participant notes
The late-arrival drill mirrored a real cutover headache. Still wish we had one more week on daylight-saving quirks.
Clear pacing; the final checklist now lives in our team wiki.