Time Series: Pick One Path
Time series appears in three places in this curriculum. You do not need all three unless you are specializing in forecasting.
| Path | Location | Best for | Depth |
|---|---|---|---|
| A. Module | 15-time-series-analysis | Theory + methods (ARIMA, Prophet, LSTM) | Full module + 10 exercises |
| B. Project | Intermediate project 6 | Portfolio piece after Stages 3–4 | Applied brief + starter.py |
| C. Advanced | Advanced project 3 | End-to-end pipeline + deployment narrative | Capstone-style brief |
Recommended choice
| Your goal | Pick |
|---|---|
| Data scientist / analyst needing forecasting | Path A (module), then optionally Path B for portfolio |
| Already know ARIMA basics; want a GitHub project fast | Path B only |
| ML engineer building production forecasting | Path A → Path C |
| Not doing forecasting roles | Skip all three, use Stage 6 time on CV/NLP or tabular projects |
Prerequisites (all paths)
- Module 01: Pandas, dates, basic plotting
- Module 05: train/test discipline (no random split on time!)
- Path A/C deep learning sections: Modules 09–10
Anti-patterns
- Doing module 15, intermediate project 6, and advanced project 3 back-to-back without new datasets or methods
- Random
train_test_spliton time-ordered data (see module 15 evaluation section) - Starting module 15 before you can load and resample a datetime index in Pandas
Related content elsewhere
- Date/time wrangling: 01 → Working with Dates & Times
- TS features in tabular ML: 07-feature-engineering
- TensorFlow TS intro: 10-deep-learning-frameworks
Back to: START-HERE.md · LEARNING_ROADMAP.md