Essential session storage is required to sign in. Optional analytics remain off unless you allow them.
Build a properly validated supervised model and compare a few algorithms. Use a scikit-learn PIPELINE + ColumnTransformer so preprocessing is fit on train only and applied identically at inference (preventing leakage). Use k-fold CROSS-VALIDATION for model selection and hyperparameter TUNING (grid/random search) — never touch the test set during tuning. Report a full evaluation (confusion matrix + precision/recall/F1/ROC-AUC for classification, or RMSE/MAE/R2 for regression) on a final held-out test set, discuss overfitting vs underfitting (bias-variance), and interpret which features mattered. Make it reproducible (seeds, environment).
k-fold CV for selection/tuning; final test set untouched until the end; correctly separates parameters from hyperparameters.
scikit-learn Pipeline/ColumnTransformer so preprocessing fits on train only; the same transforms apply at inference.
Appropriate metrics with a confusion matrix/ROC where relevant; bias-variance discussed; feature importance interpreted.
Seeds + documented environment; the notebook/script reruns to the same result.
Use this README structure to make the work reviewable.
DevPath fetches a read-only, size-capped snapshot. Submitted code is never executed.
# A validated model with tuning and a clean pipeline ## Goal Build a properly validated supervised model and compare a few algorithms. Use a scikit-learn PIPELINE + ColumnTransformer so preprocessing is fit on train only and applied identically at inference (preventing leakage). Use k-fold CROSS-VALIDATION for model selection and hyperparameter TUNING (grid/random search) — never touch the test set during tuning. Report a full evaluation (confusion matrix + precision/recall/F1/ROC-AUC for classification, or RMSE/MAE/R2 for regression) on a final held-out test set, discuss overfitting vs underfitting (bias-variance), and interpret which features mattered. Make it reproducible (seeds, environment). ## Acceptance criteria - [ ] Validation rigor: k-fold CV for selection/tuning; final test set untouched until the end; correctly separates parameters from hyperparameters. - [ ] Leakage-safe pipeline: scikit-learn Pipeline/ColumnTransformer so preprocessing fits on train only; the same transforms apply at inference. - [ ] Evaluation depth: Appropriate metrics with a confusion matrix/ROC where relevant; bias-variance discussed; feature importance interpreted. - [ ] Reproducibility: Seeds + documented environment; the notebook/script reruns to the same result. ## Verification - [ ] Document installation and run commands. - [ ] Record automated checks and their exact commands. - [ ] Add representative output, screenshots, or a short demo where useful. - [ ] Confirm failure paths and known constraints. ## Decision log ### Decision title - Context: - Choice: - Alternatives considered: - Tradeoffs: ## Evidence - Link each rubric criterion to the file, test, or artifact that demonstrates it. ## Known limitations - List what is intentionally out of scope and what should be improved next.