Time Series Forecasting

Curriculum

1. 01Forecasting for DecisionsWhat a forecast is for, how the decision sets the horizon and the level of detail, and a first look at four years of daily demand at Kolanut's Lagos depot.20 min
2. 02Trend, Seasonality and EventsBreak a series into its building blocks (trend, weekly and yearly seasonality, paydays, holidays and promotions) by measuring each one directly from the data.15 min
3. 03Baselines and Honest TestingTest forecasts the only honest way (on a later period the method never saw), measure them with WAPE, and set the simple baselines any real method has to beat.15 min
4. 04Regression with Calendar FeaturesForecast with a regression model built from what you know about future dates (weekday, month, paydays, Eid, promotions), and learn why lags shorter than the horizon are off-limits.20 min
5. 05Promotions, Events and Structural BreaksMeasure what promotions and Eid really add (including the dip after a promotion), and handle a structural break like a price rise so the forecast doesn't keep predicting a world that's gone.25 min
6. 06Backtesting over TimeTest a forecasting method on many past periods with rolling-origin backtesting, so one lucky or unlucky test period doesn't decide which method you trust.15 min
7. 07Uncertainty and OrderingPut a range around a forecast, check whether the range is honest, and turn forecast plus uncertainty into an order quantity with safety stock for the service level the business wants.25 min
8. 08Forecasting Many SeriesApply one tested method to every product, compare accuracy across them, and decide where human review is still needed.15 min
9. 09Final ProjectPlan your final project, a tested forecasting and ordering system for all six products at Kolanut's Lagos depot, and start by backtesting a product other than the malt drink.20 min
10. Final assessmentPass it (60% or more) to earn the Time Series Forecasting course completion badge.