✅ Deliverables
- Problem statement: what you’re predicting, who uses it, and why it matters.
- Dataset + data dictionary: sources, target definition, leakage checks.
- Baseline model: simple regression baseline and error analysis.
- Final model: improved features + validated performance.
- Business translation: how prediction error becomes dollars / operations impact.
🎯 Recommended evaluation
- Core metrics: MAE, RMSE, \(R^2\)
- Business-aligned metric: e.g., under‑forecast penalty vs over‑forecast penalty
- Validation: time-aware split if the data is temporal; otherwise stratified sampling by key segments
If your data is time-ordered, avoid random shuffles; use a chronological holdout and/or rolling validation.
🧩 Suggested project themes
- Retail: weekly store demand forecasting
- Hospitality: nightly occupancy / ADR prediction
- Operations: lead time / throughput / backlog prediction
- Marketing: incremental revenue prediction from campaigns
🚀 Start here
If you want a guided refresher before building the project, start with the interactive Module 2 storyline and then move to the PDF notes.
Tip: keep your final artifact reproducible (same split, same seed, same features). If you share code, include a README with run steps.