"Keep or let go?"

• Logistic regression — predicts the probability of a yes/no outcome. Coefficients are interpretable as log-odds.
• Decision trees — produce human-readable if/then rules. Greedy splits on the feature that best separates the classes. Easy to over-fit.
• Random forests & ensembles — many imperfect trees averaged together usually beat one "best" tree. Less interpretable, more accurate.
• Confusion matrix — true positives, false positives, true negatives, false negatives. The single most important table in supervised learning.
• Precision vs. recall — depending on the business cost, you optimize for one or the other (or balance both via F1).
• ROC & AUC — model quality across all thresholds; AUC of 0.5 is a coin flip, 1.0 is perfect.
• Train/test split & cross-validation — never trust a model on data it has seen.
• Class imbalance — if 99% of cases are "no," accuracy is meaningless. Use balanced metrics.

• Classification Basics (Week 6)Foundations: what classification is, when to use it, evaluation metrics.
• Decision Tree Logic (Week 6)How trees split data; gini vs. entropy; pruning; reading a tree.
• End-to-End Pipeline Activity (Week 7)From raw CSV to a trained, validated, defensible classifier.
• Saving Lives with BizML — Stent FailureReal-world stakes: when a classifier influences medical-device decisions.

• Statistics & ML Cheat SheetOne-page reference: distributions, hypothesis tests, ML terminology.
• When Models Go Wrong — Case StudiesFive cautionary tales of classifiers that misled their users.
• Predictive Analytics & Data Mining HandbookReference textbook chapters covering this entire module.

The Week 6 in-class activity walks you through building a decision tree on a real loan dataset, then defending your splits to a partner.

• Decision Tree Activity — Student WorksheetFill-in-the-blank: build a tree, justify your splits, predict outcomes.
• Loan Application Dataset~500 loan applications with approve/deny outcomes for the worksheet.

• Predicting Customer Churn — Logistic Regression & Decision TreeFull pipeline: load, EDA, split, train both models, compare results.
• Week 7 — ML in HealthcareApply classification to a healthcare dataset; discuss bias and fairness.
• Random Forest NotebookBuild a random forest; tune hyperparameters; compare against the single tree.

• Telco Churn DatasetClassic customer-churn dataset for the logistic regression walk-through.
• Loan Application DatasetUsed in the Week 6 in-class activity.
• Stent Failure DatasetMedical-device classification — high-stakes, asymmetric error costs.

A small medical-device company wants to predict which stents are at high risk of failure post-implantation. False negatives are catastrophic; false positives mean unnecessary surgeries. Build, evaluate, and present a classifier — once for executives, once for the technical team.

• Stent Failure Dataset~2,000 procedures with patient features and failure labels.
• Executive PresentationSample executive-audience deck — what to lead with, what to leave out.
• Technical ReviewSample technical-audience deck — methodology, validation, caveats.

Same problem, enterprise tool. Use SAS Viya Model Studio to build a healthcare-classification pipeline visually, with governed data prep and reproducible scoring.

• SAS Model Studio InstructionsStep-by-step guide to the SAS Viya pipeline.
• Case Study BackgroundBusiness context for the healthcare classifier.
• Sample Business PresentationHow to frame the model for the hospital leadership audience.
• Sample Technical PresentationHow to present the same work to the data team.

Three games that drill the "which model wins?" muscle that classification really is about.

• Analytics ConsultantPlay consultant: read the scenario, pick the right method, defend it.
• Model MatchmakerMatch models to problems based on data shape and outcome type.
• Prediction ShowdownCompare candidate models head-to-head on the same dataset.

Two of the end-of-course cases use classification as their primary method:

• ShieldScore — Auto insurance fraud detection · classification + deployment
• SkyBridge — Aircraft engine predictive maintenance · classification + deployment

If you want to push your skills, work through one of these after finishing the topic materials.