Skip to the content.

Class project

← Back to course home · Schedule · Syllabus · Policies

Beginning in the second week of class, groups of about four students work together on a machine learning project throughout the semester, using one of several real-world public policy problems. Each week, every group submits a project assignment. In addition to connecting the readings and discussion topics to the policy domain, these updates give you a chance to get input and feedback and iterate.

Project Descriptions

This semester, we have set up two projects

  1. Reducing Jail Rebookings through Proactive Mental Health Outreach
  2. Supporting Advocacy for Civil Rights by Prioritizing State Bills that are likely to Pass

⚠️ Data security. Project data is sensitive and must remain in the secure computing environment provided for the course. See the data security policy — violations result in automatic failure of the class.

Deliverables at a glance

# Deliverable Due Weight
1 Project proposal Tue, Sep 15 10%
2 Peer reviews of three proposals Tue, Sep 22 5%
3 Weekly progress updates Most Tuesdays 25%
4 Final presentation Thu, Dec 3 10%
5 Final report and code Wed, Dec 9 25%

20% of your grade is class attendance and participation and 5% is weekly feedback forms — see policies.


1. Project proposal

Submitted as a group, 4–5 pages not including figures, tables, or references. It should cover the project scope and preliminary descriptive statistics about the entities in your dataset, and include the provided scoping worksheet as an appendix.

Useful reference: the Data Science Project Scoping Guide.

2. Proposal peer reviews

Each student reviews three other groups' proposals. Reviews should be constructive and specific: is the problem well scoped, is the proposed formulation the right one, what would you imorove?

3. Weekly Project Assignments

Short update assignments, generally due Tuesdat, that guide the check-in discussions. These typically take the form of filling results or modeling details into a handful of template slides. Over the semester these build up the technical core of the project:

Updates are graded for completeness and correctness. We expect this work to be iterative: errors identified in one week's update that are corrected by the next week results in revision of the previous score up to 80% of the total possible.

4. Final presentation

15 minutes plus 3 minutes for questions. The final presentation should be geared toward the relevant decision makers for your project: an overview of the problem and approach, your results, policy recommendations, and limitations of the work.

6. Final report

Approximately 10 pages, accompanying the final presentation. It should include:


Project iteration targets

The point of the weekly cadence is to have a working end-to-end system early and improve it, rather than building the pieces separately and integrating at the end.

Iteration Weeks Focus
1 5–6 End-to-end shell — the simplest possible full pipeline
2 7–8 Feature development
3 9–10 Models and evaluation
4 11–12 Interpreting the models
Final 13–14 Final model choice, disparities, impact

Templates and worksheets

Scoping worksheets, update slide templates, and report templates are in the project/ directory.