Foresight Editor-in-Chief Len Tashman's Preview of the Fall 2019 Issue
This 55th issue of Foresight opens with an article from Phillip Yelland, Zeynep Erkin Baz, and David Serafini of the Data Science/AI team at Target: Forecasting at Scale: The Architecture of a Modern Retail Forecasting System. The challenge of scale at Target is the requirement for nearly one billion forecasts per week, demanding a careful balancing act between statistical modeling, software engineering, and business practice. The team also reports the need to weigh considerations of forecast accuracy against model explainability.
We found that while some statistical forecasting models might achieve superior forecasting accuracy, it was generally at the expense of explainability; we chose models that represented the best compromise between these twin considerations.
The lead author, Phillip Yelland, talks about his role in our Forecaster in the Field interview.
One of Foresight’s most visionary authors, Niels van Hove, looks ahead to the emergence of Autonomous or “Lights Out” Supply-Chain Planning and identifies the essential requirements for this new technology to take hold.
Autonomous car development is happening in most of the major car companies with the promise of self-driving cars everywhere within a few years. Since autonomous vehicles already exist in the supply chain, why should we not strive to achieve autonomous supply-chain planning? However, to truly achieve “lights out” planning, eliminating the human planning role, the problem-solving and decision-making capabilities of the system must improve upon human reasoning, judgment, and creativity so as to resolve impediments to execution of solutions.
Niels’ article is followed by a Commentary from Stefan de Kok, who argues that the autonomous supply chain is not going to happen if the separate horizons of operational, tactical, and strategic planning continue to operate in isolation from each other. We not only need to close the loops in planning but be able to rapidly respond to changes through autonomous replanning.
Our book review is of the hot-off-the-press Forecasting: An Essential Introduction, by Jennifer Castle, Michael Clements, and David F. Hendry. Reviewer Mike Gilliland, Foresight’s Editor for Forecasting Practice, observes that while business forecasting and economic forecasting “overlap in many ways, this book exposes some significant, fundamental differences that make economic forecasting an even more vexing challenge.”
Autonomous transportation and autonomous planning are two of the applications of futuristic technology and artificial intelligence. This edition of Foresight concludes with Dan Philps’s article Continual Learning: The Next Generation of Artificial Intelligence, which describes the emergence of autonomous machine learning and highlights how Dan’s team applied CL to investment decisions. The article addresses how automated machine learning might mitigate forecasters’ concerns with the complexity of AI.
While ML is generally perceived to be complex, it can actually reduce complexity in model development by avoiding human behavioural biases and by automating intermediate steps. In addition, if complexity serves to encompass a richer variety of information and if learning from this information can be automated, complexity is a price worth paying.
COMING UP RIGHT AWAY:
THE 2019 FORESIGHT PRACTITIONER CONFERENCE
November 13-14 are the dates for this year’s Foresight Practitioner Conference, where in beautiful Chapel Hill, North Carolina, a distinguished group of speakers will offer their insights into Artificial Intelligence: The Hype and the Promise for Forecasting and Planning. Hope to see you there!
Find details on the program and registration at: https://foresight.forecasters.org/2019-conference/.
