Your VP of Sales projected $4.1M in monthly revenue by December 2026 — 18% year-over-year growth. The data engineering team found bugs in the original data that inflated recent figures. They've handed you clean data. Your job: figure out what the numbers actually say.
Open starter/starter.ipynb and work through all four phases.
pandas>=2.0
numpy>=1.24
matplotlib>=3.7
statsmodels>=0.14
darts>=0.41
scipy>=1.11
scikit-learn>=1.3
pip install pandas numpy matplotlib statsmodels darts scipy scikit-learnWork through the notebook in order:
- Phase 1: Baseline Forecasts — naive, moving average, linear trend
- Phase 2: Classical Models — ARIMA, SARIMAX with diagnostics and prediction intervals
- Phase 3: Modern Models — N-BEATS (neural) and Chronos-2 (foundation model) via Darts
- Phase 4: Comparison and Recommendation — comparison table + written recommendation
- Completed notebook with all phases implemented
- Comparison table showing all models, December 2026 forecasts, prediction intervals, and accuracy metrics
- At least three charts: baseline comparison, classical model forecasts with intervals, modern model forecasts
- Written recommendation (500-1000 words) answering:
- What is the realistic range for December 2026 revenue?
- What should the company plan for instead of $4.1M?
- Which forecasting approach would you recommend for ongoing use, and why?
data/revenue.csv— 60 months of monthly revenue (January 2021 through December 2025), cleaned by the data engineering team