Modules

Welcome to the Modules section. Over the course of the semester, you’ll build one forecasting model and progressively improve it — adding smarter components, external context, and production-aware robustness with each module.

1 The course arc

Module 1 Decomposition baseline Module 2 ETS & ARIMA filters Module 3 Exogenous variables Module 4 Robustness & scale STL + SNAIVE + Drift STL + ETS / ARIMA ARIMA(y ~ xreg) Prophet + ensembles Understand the series Smarter trend/cycle External context Production-ready Ch. 2–5 Ch. 8–9 Ch. 7, 10 Ch. 12–13
TipHow to navigate

Use the sidebar to move lesson-by-lesson through each module. Each lesson includes explanations, worked examples in R, and exercises to reinforce the ideas.

2 Prerequisites and setup

Make sure you can open an .Rproj, run an {r} code chunk, and install packages in R before starting. If you’re unsure, Module 1 — Introduction covers the course workflow and expectations.

3 Course map

3.1 Module 1: Forecasting models based on decomposition methods

Build the fundamentals: time series data structures, transformations, decomposition, and your first complete forecast. You’ll establish a benchmark model that every later module will try to beat.

Once you have a working baseline, Module 2 replaces its naive components with smarter statistical filters →

3.2 Module 2: Adding ETS and ARIMA filters

Replace the benchmark components with exponential smoothing and ARIMA models. Learn to identify and correct non-stationarity, and combine decomposition with full statistical models.

Your model now handles trend and autocorrelation well — Module 3 adds external information from outside the series →

3.3 Module 3: Regression and exogenous variables

Incorporate external drivers and handle real data challenges. You’ll move from clean regression setups to dynamic models that combine regression with ARIMA errors, and close with Prophet as a practical industry tool.

With external variables in the model, Module 4 focuses on making everything robust and production-ready →

3.4 Module 4: Forecasting at scale

Handle complexity: multiple seasonal patterns, uncertainty quantification, and forecast combinations. The final module ties everything together into a complete, production-aware workflow.

  • 4.1 Complex Seasonality (coming soon)
  • 4.2 Bootstrapping, Bagging & Model Combinations (coming soon)
  • 4.3 Time Series Cross-Validation (coming soon)
  • 4.4 Hierarchical & Grouped Forecasting (coming soon)

Complete Modules 1–3 before starting here.

Back to top