Activity: Write a Machine-Learning Problem Brief

Activity ID: ML-M01-A01 Estimated active time: 35-50 minutes

Goal

Create a clear problem brief before any model is trained.

Case options

Choose one case.

Case A: Learner Support

An online course team wants to help learners who may stop during Module 1. The team can send a short optional support message, but mentors have limited time.

Case B: Support-Ticket Priority

A small education company receives support tickets from learners. The team wants to find tickets that may need a faster response.

Case C: Practice Recommendation

A course platform wants to recommend extra practice resources to learners who may benefit from more examples.

Instructions

1. Pick one case and write the decision it supports, in one sentence.
2. Name the unit of prediction: one learner, one session, one transaction.
3. Name the target and say exactly when it becomes known.
4. List the features available at prediction time, and separately list anything recorded later.
5. State what a false positive costs and what a false negative costs. They are rarely equal.
6. Write the success criterion: what must this beat to be worth deploying?

Quality checklist

• [ ] The decision is named, not just the prediction.
• [ ] The unit of prediction is stated explicitly.
• [ ] Every listed feature is genuinely available before the target is known.
• [ ] The two error costs are described separately.
• [ ] Success is defined against a named baseline, not in the abstract.

Example opening

Weak:

Predict urgent support tickets.

Better:

Within 10 minutes of a support ticket arriving, predict whether the ticket is likely to require same-day human response, so the support lead can review high-priority tickets first.

Submission format

Submit the completed table and short notes as plain text or Markdown. Do not include real names, emails, phone numbers, employer data, private learner data, or sensitive information.