This book is licensed for the personal use of the buyer. Please do not share or resell the file. Short quotations with attribution are welcome.
The information in this book is provided for education. It is not legal, financial or other professional advice. Product and company names are used only to describe them and belong to their owners.
This is a free sample: the introduction and first chapter.
IntroductionIntroduction
Forecasts guide decisions everywhere: how many staff to schedule, how much stock to order, how much cash will be available, how many servers to run, how many patients will arrive. A good forecast saves money and prevents crises. A bad one, especially a confident bad one, does the opposite.
This book teaches forecasting that is both useful and honest. It starts with exploring time series and understanding trend, seasonality and noise, then builds strong baselines that every forecast must beat. It covers exponential smoothing, ARIMA-style models, regression with features, machine learning and forecasting many series at once. It gives special attention to evaluation with rolling backtests, to prediction intervals and to the events that break forecasts: holidays, promotions and sudden shocks.
Worked examples are small enough to check by hand.
How the book is organised
Chapters 1 to 8: what forecasting is, time series data, components, exploration and preparation.
Chapters 9 to 15: baselines and classical models.
Chapters 16 to 23: features, machine learning, many series, hierarchies and special demand patterns.
Chapters 24 to 31: evaluation, metrics, backtesting and uncertainty.
Chapters 32 to 45: practice: events, shocks, judgement, communication, code, deployment, monitoring and applications in several sectors.
Chapters 46 to 51: worked trend smoothing, anomaly detection, nowcasting, intervention effects, short histories and data-quality checks.
Chapters 52 to 58: case studies, common mistakes, building a practice and a checklist.
Chapter 1What Forecasting Is For
A forecast exists to support a decision. Start there.
Decision
Forecast needed
Horizon
Granularity
Staff rota
Customer arrivals
1–4 weeks
Per hour
Stock order
Unit demand
Supplier lead time
Per product per store
Cash planning
Receipts and payments
13 weeks
Per week
Capacity planning
Peak load
6–24 months
Per month
The decision determines the horizon, the level of detail, the accuracy needed and whether a range or a single number is required.
You have reached the end of the free sample. The full book continues with 57 more chapters, plus the appendices.