October 23 webinar: how demand planning will benefit from machine learning


Applying machine learning approaches to forecasting is an area of great research interest. Progress is being made on multiple fronts, for example:

  • In the M4 Forecasting Competition, completed earlier this year, the top two performers utilized machine learning with traditional time series forecasting methods. At the link you'll find full details on the rules, the 100,000 time series used in the competition, and results.
  • At the SAS Analytics Experience 2018 in San Diego last month, SAS customer Kellogg's along with SAS partner First Analytics reported on a machine learning approach to guide overrides of statistical forecasts to improve accuracy. (Register here to watch their Tuesday September 18 presentation "How Machine Learning Boosts Statistical Forecasting for Better Demand Planning at Kellogg’s" on demand.)

And on Tuesday October 23, noon-1:00pm ET, SAS is presenting a live webinar:

How Demand Planning Will Benefit From Machine Learning

My colleagues Varun Valsaraj, Becky Gallagher, and Charlie Chase present their approach to enhancing business forecasting with machine learning. Their novel approach helps demand planners identify which forecasts should be manually adjusted, and provides guidance on the direction and size of the adjustment. This method shows promise at reducing the number of overrides made by planners (saving time), and improving the quality of the overrides (increasing Forecast Value Added).

Register here for the free live webinar, and check back for the on-demand recording.


About Author

Mike Gilliland

Product Marketing Manager

Michael Gilliland is a longtime business forecasting practitioner and formerly a Product Marketing Manager for SAS Forecasting. He is on the Board of Directors of the International Institute of Forecasters, and is Associate Editor of their practitioner journal Foresight: The International Journal of Applied Forecasting. Mike is author of The Business Forecasting Deal (Wiley, 2010) and former editor of the free e-book Forecasting with SAS: Special Collection (SAS Press, 2020). He is principal editor of Business Forecasting: Practical Problems and Solutions (Wiley, 2015) and Business Forecasting: The Emerging Role of Artificial Intelligence and Machine Learning (Wiley, 2021). In 2017 Mike received the Institute of Business Forecasting's Lifetime Achievement Award. In 2021 his paper "FVA: A Reality Check on Forecasting Practices" was inducted into the Foresight Hall of Fame. Mike initiated The Business Forecasting Deal blog in 2009 to help expose the seamy underbelly of forecasting practice, and to provide practical solutions to its most vexing problems.

1 Comment

  1. As you've pointed out, the top 2 performers in the M4 competition weren't using pure ML, but a statistical combinations aided by ML (I think that's probably the easiest way to describe it). It is also worth mentioning that the top 6 performers' results weren't that significantly different and that the other 4 in the top 6 didn't use ML. As a whole in the M4 competition, ML didn't add meaningful value ... meaning that - right now, there is little benefit for large scale forecasting consumers to jump on thee bandwagon without a proper, out of sample proof of concept that compares ML results to the results of both their current system as well as forecasting benchmarks used in the competition. At least, this is the view that's shared by many retail forecast scientists I'm in contact with.

    That doesn't mean that ML can't or won't work well for a large group of customers - solution providers have access to large numbers of data sets that allow them to make specific adjustments and improvements to the algorithms they employ and this 'advantage' isn't necessarily portrayed in the M4 results.

    I'm pretty sure ML will improve drastically over the next few years and it's really up to the end user to be aware of these potential limitations.

Leave A Reply

Back to Top