Foundation project·60-90 minutes

Foundation Project: Evidence Before Models

Audit a small tabular dataset, create a leakage-safe summary, and decide what can be claimed before training a complex model.

Scenario

A study-planning team wants to predict course completion from hours studied and prior quiz results. You must determine whether the sample supports a modeling experiment and establish a transparent baseline.

You will demonstrate

Define one row, target, prediction time, and error costs.
Compute training-only summaries and a deterministic baseline.
Separate implementation evidence from generalization evidence.

Project evidence

Complete each deliverable.

Deliverable 1

Write the data contract

Define every field, its unit, availability time, and allowed range.

Evidence: A compact data dictionary and target definition.

Deliverable 2

Defend the split

Choose random, grouped, or chronological evaluation and explain how it mirrors use.

Evidence: A split rationale with one rejected alternative.

Deliverable 3

Build the baseline

Implement the fixed threshold baseline in the code lab and report confusion counts.

Evidence: Passing tests and a four-count confusion matrix.

Deliverable 4

Make a bounded recommendation

Recommend continue, revise, or stop without extending beyond the sample.

Evidence: A six-sentence decision memo with the next evidence to collect.

Runnable Python lab

Build and audit a threshold baseline

Compute confusion counts and cost for a transparent pass-risk rule.

Test output

Run the code when you are ready. The first load downloads the in-browser Python runtime.

Code runs only in a time-limited browser worker. It is not sent to the IntuitiveAI server. Do not paste secrets.

Expected result

{'TP': 4, 'FP': 0, 'TN': 4, 'FN': 0, 'cost': 0}
ALL TESTS PASSED

Project defense

Which result most directly invalidates the evaluation?

Assessment rubric

Definitions are precise enough for another learner to reproduce.
Held-out information does not influence fitted values.
Every claim points to an observable result.
Limitations and next evidence are explicit.

Useful references

Model Cards for Model ReportingA reporting framework for intended use, evaluation context, limitations, and ethical considerations.

Project completion gate

Complete all deliverables, pass the automated code checks, and defend the key decision.

Deliverables 0/4Code pendingDefense pending