Data Science Book Review: The Business Forecasting Deal | Data Mining Research

If you’re new here, you may want to subscribe to my RSS feed. Thanks for visiting!

data-science-book-review-the-business-forecasting-deal-data-mining-research Data Science Book Review: The Business Forecasting Deal | Data Mining ResearchThere are not many books on forecasting. Even less good books. The Business Forecasting Deal, by Michael Gilliland, is one of them. The book is among the rare ones providing real practical advices for professionals. No theory, no equation, no code. The book focuses on typical forecasting challenges, misconceptions and issues when deploying processes into production. The main point of the author is not to spend to much effort on forecasting, but rather on implementing and deploying those forecasts into the company.

Gilliland’s book, targeted to business users, is a complement to other technical books such as Forecasting: Principles and Practices (Hyndman and Athanasopoulos) and practical ones such as Future Ready (Morlidge and Player). Although practical, The Business Forecasting Deal discusses very specific issues in forecasting such as worst practices and forecasting without history.

The main concept of the book is the so called Forecast Value Added (FVA). The point of Gilliland is that forecasts are useful only when they add something to a naive prediction (last value, average, etc.). If your forecast isn’t better than a naive one, why spend time and resources on it? Common measures such as MAPE (Mean Absolute Percentage Error) aren’t enough as they don’t provide an indication of how much better/worse than a naive forecast you are.

Gilliland always mentions MAPE as a limited way of measuring forecasting accuracy. However, practitioners should know a better way of measuring accuracy: Mean Absolute Scaled Error (MASE). Advised by Hyndman (article here), it implicitly includes a naive forecast within the equation to create a useful accuracy metric. For me, it looks like FVA is another name for existing metrics such as MASE. Hopefully, the FVA is more than just an equation. It’s a full methodology allowing to reduce inefficiencies in the overall forecasting process. The book concludes with a large Q&A section which is often repetitive of the book content.

The Business Forecasting Deal is an excellent book containing plenty of sound advices for forecasting practitioners. Gilliland did an excellent job in summarizing challenges and issues in forecasting, and providing related solutions. In complement to an introductory business book and a technical one, it will allow you to integrate forecasting within your company and gain the best of it.

what-could-big-data-mean-for-debt-management-data-mining-research Data Science Book Review: The Business Forecasting Deal | Data Mining Research




Source link

Previous Use your phone, or Alexa, to control Bosch's 800 Series smart appliances
Next Mujjo leather iPhone 7 Case Review