Smoothing 6: Winter's exponential smoothing

Winter's (Holt-Winter's) exponential smoothing is a popular data-driven method for forecasting series with a trend and seasonality.
This video supports the textbook Practical Time Series Forecasting.
www.forecastingbook.com
www.galitshmueli.com

[@bartvandenberg325]

All three videos on exponential smoothing are excellent

[@danielschauder9359]

This series on exponential smoothing is really outstanding. I felt the concepts were very clearly explained, and the impact of each tuning parameter was fully described. I really appreciated the real-world examples and the visualizations.

[@federicogonzalez3438]

Excellent videos and explanation! Thank you very much, you did a fantastic job with this channel!!

[@herogenius0505]

This video and the rest 2 in the ES series are super useful and insightful. I fully understand them now

[@apttutorial]

Thank you, ma'am. You explained it very well.

[@folarinlonge5566]

Thank you. This was helpful

[@mingg.5546]

Great video and explanation. Thanks Dr Shmueli. However, I don't quite get the 3:06 where we need to de-seasonalise the value of Y. Does this mean that in the end we need to add the seasonality back for forecasting? The video didn't mention the need for this step though. How is this related to the formula used in the Coke example Ft+k = (Lt + kTt)*St+k-m ? Thanks for any tips!

[@thagoonlprusiriyodlpru4174]

Thank you so much

[@uwen.13]

Thanks for the video and also the book, which I bought recently. There is one thing I do not get: What is the real difference between smoothing and a regression with time dummies? In a regression, the coefficients for the time dummies are chosen in such a way, that the SSE is minimized. In a smoothing model (like Holt-Winters with additive components), the coefficients for the smoothing components are basically also selected in such a way, that the error terms are minimized. Therefore - in a linear regression and a smoothing model with additive components - the same results should appear. Right?

[@ZhijianCai]

Great video! But what if the seasonal period is unknow or time-varying? how can i estimate it?

[@johannaw2031]

Can we combine exponential smoothing with regression?