# Term Project Manual — CSCI 49395

Big Data Analytics and Applied Privacy. This manual is the specification for the term
project. Read it in full before writing your proposal.

## Goal

Design and implement a substantial, end-to-end big-data or privacy artifact and report on it
honestly. Your project must engage a real privacy or scale challenge — not just call a library.
Separate what you *measured* from what you *believe*.

## Deliverables

The project is submitted in four parts, in order:

1. **Proposal** — 1–2 page PDF: problem, dataset(s), method, privacy considerations, and a
   milestone plan. Must be approved before you begin implementation.
2. **Code repository** — a public or shared Git repo with a `README`, a pinned environment
   (`requirements.txt` / lockfile), and a one-command way to reproduce your results. Raw data
   stays read-only; all transforms are code.
3. **Final report (PDF)** — 6–10 pages: introduction, related work, method, results with
   effect sizes / uncertainty, a dedicated **privacy analysis**, limitations, and conclusion.
4. **Jupyter notebook** — a runnable notebook that reproduces the key results end-to-end. It
   must run top-to-bottom (Restart & Run All). See `example-project.ipynb`.

## Required format

| Deliverable | Format | Notes |
|---|---|---|
| Proposal | PDF | Title, names, references. |
| Code repo | Git URL | README + pinned env + seed; no committed secrets or PII. |
| Final report | PDF | Figures with captions; cite every dataset and claim. |
| Notebook | `.ipynb` (nbformat v4) | intro → data-load → analysis → **privacy analysis** → conclusion. |

## Rubric

| Component | Weight | What we look for |
|---|---|---|
| Proposal | 10% | Clear question, feasible scope, privacy awareness. |
| Implementation (code + notebook) | 40% | Correctness, reproducibility, code quality. |
| Privacy analysis | 20% | Explicit threat model + a real mitigation, evaluated. |
| Final report | 20% | Honest claims, effect sizes + uncertainty, limitations. |
| Presentation | 10% | Clear communication of the problem and results. |

## Timeline

Dates follow the course schedule (see the syllabus). Indicative milestones:

- **Week 4** — Proposal due.
- **Week 11** — Progress check-in (repo + partial results).
- **Week 15** — Final report, notebook, and presentation due.

## Integrity & privacy rules

- Never commit PII, credentials, or licensed/embargoed data. De-identify before sharing and
  keep raw data out of the repo.
- Report the honest result, including nulls. Preference is not the same as a measured gain.
- Cite every dataset (license + source) and every external result you compare against.