STAGE 1: Frame the decision
Concept: A prediction target must support a defined action.
Task: Define a completion target, observation window and intended support action.
Acceptance: Explain the cost of false positives and false negatives.

My assumptions:

My work / file link:

My checks and results:

What I changed:


STAGE 2: Inspect the data
Concept: Associations do not establish causes.
Task: Create a data dictionary and mark features available before the prediction time.
Acceptance: Exclude outcome-derived fields such as final certificate status.

My assumptions:

My work / file link:

My checks and results:

What I changed:


STAGE 3: Build a baseline
Concept: A complex model must earn its complexity.
Task: Specify a simple majority-class baseline and a train/validation/test strategy.
Acceptance: Do not tune decisions using the final test set.

My assumptions:

My work / file link:

My checks and results:

What I changed:


STAGE 4: Evaluate responsibly
Concept: Overall accuracy can hide class imbalance and subgroup errors.
Task: Calculate precision and recall from the sample confusion counts and discuss error review.
Acceptance: State that a tiny synthetic fixture cannot establish real-world performance.

My assumptions:

My work / file link:

My checks and results:

What I changed:
