Apply the full forecasting pipeline, exploration, decomposition, stationarity, model fitting, and backtesting, to a brand new series with its own structure: weekly data, multiplicative seasonality, and a genuine change in growth rate.
Welcome to the Capstone, the ninth and final module of the course. Every technique so far was built and tested on Cyclepath, a monthly series with a constant trend, an additive season, and a period of 12. That consistency was deliberate: it let each module isolate one new idea at a time. This module removes that safety net. You will meet Lantern & Vine, a fictional home-goods shop’s weekly unit sales, a series with its own structure, and you will run the entire pipeline on it from the very beginning, with nobody telling you the answers in advance.
Lantern & Vine differs from Cyclepath in three real ways at once: it is weekly instead of monthly, so its seasonal period is 52, not 12. Its seasonal swing scales with its growing size, a multiplicative structure, rather than staying a fixed amount. And its growth rate itself changes partway through the series, something Cyclepath never did. You will explore it, decompose it, diagnose its structure with the exact same tools Module 2 built, detect the change in its growth rate, make it stationary, identify and fit candidate models, and backtest them properly using Module 8’s discipline, all before producing a genuine forecast of the next six months.
Every number in this module is computed for real, on a newly seeded series built specifically for this capstone, using the same standard of verification this course has held to from Lesson 1 of Module 1 onward. Start with Lesson 1, meeting the series for the first time.
Complete all 5 lessons to finish the Capstone module.