Lesson 14 — Capstone Presentations & Course Wrap-Up
(end-to-end analysis, communication, and reflection)
Class plan (10-minute presentations)¶
Each group (or individual) presents for 10 minutes + 3–5 minutes Q&A.
Suggested structure (10 minutes total):
Motivation and research question (1 min)
Data (1–2 min)
Method and workflow (2 min)
Results (key visuals/tables) (3 min)
Interpretation + implications (1 min)
Limitations + responsible AI note (1 min)
What the capstone must contain (final submission)¶
Allowed scope (choose ONE main method; optional secondary analysis)¶
Choose ONE as your main method¶
Regression (simple/multiple)
Unsupervised learning (clustering and/or PCA)
Decision tree model (classification)
Text-as-data (sentiment / simple NLP)
Time series forecasting (baselines + evaluation)
Optional: you may add one secondary method briefly, but the project should remain focused and interpretable.
Expectations for quality (what “good” looks like)¶
A good capstone shows:¶
Clear question (not just a topic)
Clean data with documentation (data source note + cleaning choices)
Appropriate method with justification
Verification (sanity checks, baselines, and/or robustness checks)
Communication (good charts + clear narrative)
Responsible AI (prompt log + no fake citations + honest limitations)
Ethics/fairness awareness (at least one risk + one mitigation/monitoring idea)
Verification examples (pick 3–5)¶
missingness summary (overall and by group)
outlier check and handling rule
baseline comparison (for forecasting or classification)
train/test split that avoids leakage (especially time series)
sensitivity check (e.g., alternate variable definition, alternate k in clustering)
manual spot-check for NLP outputs (inspect misclassified examples)
Grading rubric (suggested)¶
Wrap-up discussion (in class)¶
We close the course with:
“What skills are now in your toolkit?”
“What surprised you about data work?”
“Where did AI help, and where did it mislead?”
“What would you learn next if you had 4 more weeks?”