Frame a problem, split data correctly, choose metrics, detect overfitting, and compare against a baseline.
For a small example, predict whether six students pass when five already pass. Compare the model with the majority baseline; accuracy alone exposes that a constant predictor already scores highly.
Fraud is rare and false blocks are costly. Choose time-aware splits, report precision-recall and cost at a threshold, then inspect errors by customer segment.
Initializing Interactive Playground...
Predict how one control changes the ML Workflow & Evaluation result.
The starter deliberately follows a tempting but flawed assumption: Tuning on the test set
Check: The browser checks pass and you can name the failure mechanism before opening the explanation.
Failing starter lab
Implement confusion counts for fixed probabilities and compare a default threshold with a recall-oriented threshold.
Run the incomplete starter before opening hints. Use the first wrong output or failed check to form a repair hypothesis.
Explanations after each answer · 80% to pass · ML Workflow & Evaluation
Chapter 1 of 3 · Core
State the user decision, prediction unit, target, prediction time, available inputs, and cost of errors. Then create a simple baseline: the mean for regression, the majority or prevalence-informed rule for classification, or a domain rule already in use. A candidate must justify its complexity by improving a relevant outcome.
Choose metrics before inspecting candidate results. For classification, accuracy counts every decision equally, precision asks how often positive predictions are correct, recall asks how many actual positives are found, and calibration asks whether predicted probabilities match observed frequencies.
Key equations
Machine Learning Fundamentals: Cross Validation
StatQuest
See why one train-test split is fragile and how cross-validation gives a more reliable estimate.
Primary sources and free technical reading for ML Workflow & Evaluation.
Free reference for preprocessing, model selection, metrics, and classical ML algorithms.
Free, practical guidance on baselines, pipelines, metrics, and production ML iteration.
Research paper proposing transparent documentation of model performance and limitations.