Business Forecasting

CHAPTER 3Forecasting Methods: Modeling, Selection, and Monitoring

Artificial intelligence and machine learning have dominated the attention of researchers and practitioners the past five years. Yet there has continued to be important new work in the more traditional areas of forecasting methodology. This chapter shares over a dozen compelling articles and commentaries covering advances in these traditional areas.

The chapter begins with a selection on time-series basics from the Kolassa and Siemsen book, Demand Forecasting for Managers. Then, a taxonomy of forecasting problems to help focus efforts on the right data and methods is offered for each type of problem.

Judgment plays a well-recognized role in the override of computer-generated forecasts, but this role is less well recognized in the selection of forecasting models. Ground-breaking research by Fotios Petropoulos is shared, showing that automatic model selection can be improved upon by judicious use of judgment. Multiple commentaries explore the implications of these findings.

Paul Goodwin next contributes two pieces on the more familiar role of judgment in forecasting. The first is a selection from his book Forewarned: A Skeptic’s Guide to Prediction, and the second, a summary of the most recent research on this topic.

The next two articles deal with topics worthy of the forecaster’s attention: the concept of probabilistic demand planning, and the use of prediction markets for corporate planning. While neither approach ...