Forecasting (14): Holt Winters' method

[@ramaachavan2311]

Wonderful. Will go through this again

[@RESEARCHHUB]

Feel free to follow the course at researchhub.org/course/forecasting-in-excel/

[@hydra4454]

god bless you and your family

[@shinerajukappil6295]

Expecting more videos on forecasting

[@RESEARCHHUB]

Coming soon.

[@humanspider4419]

thank you

[@adittyasuhendra326]

Thankyou so much sir you help me 🙏

[@chinomnsodaniel7234]

I really need to do all this using spss

[@pathumsameera1788]

Excellent

[@shinerajukappil6295]

Dear sir, u have mentioned that u missed to add constraints in solver to optimise the parameters. So does that affected the minimised values that given in the results.

[@RakeshKumar-wc2ng]

Dear Researcher, while initializing the Trend component, should not be multiple by Beta value as well?

[@TheDarkhorse8888]

I have data that is very trend oriented to where the seasonality at the beginning of the year is always higher than at the end of the year. The issue, is that in the forecast period, the future, this causes a spike at the start of each year far beyond what is reasonable. Is there some way to moderate this effect? In the test data, this effect is moderated by the level which automatically adjusts to a realistic amount. However, in the forecast period there is no level moderation to moderate it. How can we resolve this? Also, it would also seem that the trend and the seasonality of the data are overcompensating for the trend by double counting it.

[@RESEARCHHUB]

You should consider univariate regression approach following seasonal decomposition.

[@memoonazahidi6093]

or can we add another factor except seasion, trend ,and level?in order to modify the model...who's factor we can use?kindly suggest me the 4th factor please

[@abdulazizalsakkar6991]

What interpretation can we make from gamma's optimal value being 0 ?

[@abdulazizalsakkar6991]

Does that mean there is no effect of seasonality ?

[@RESEARCHHUB]

Although the first part of seasonal equation will become 0, it would mean that you are using the full seasonal value from previous season because then 1-Y, will be 1 which will be multiplied with S_t-m.

[@abdulazizalsakkar6991]

@@RESEARCHHUB Oh I see, thank you.

[@m-bh2fb]

Spend some good time with your maternal aunt

[@johtinann7483]

when calculating the Yt+1 at the first, why we use the St value from last year's Feburary instead of January?

[@florianewers8069]

Thank you very much for your superior video. I now understand the calculation a lot better. I have one problem though: If i compare your calc with the builtin excel forecast function the forecast func will give me completely different results for the future. Even if i use the a b and c excel used. Does excel not use ETS?? Or is it using a different formula??

[@RESEARCHHUB]

Hi, ETS can have several different formulation. Pls see otexts.com/fpp2/ets-forecasting.html

[@JoeJoe-tf8uj]

Thank you for the great video! I have a question. How to initiate st when use additive method?

[@memoonazahidi6093]

how can we modify holt winter moder by adding weighted function or by intercept term?kindly guide me

[@RESEARCHHUB]

Thank you for this comment. I am making a video on this now, will share tonight or tomorrow.

[@MAB-05]

how do I know that the fluctation is high enough?

[@ashutoshchavan812]

Can I get any simple sample code for holt winters for forecasting time series in python?

[@RESEARCHHUB]

We will soon publish forecasting videos using R software, Python is in our agenda but will take some time.

[@syedimranali9]

Hi. Can you make some video on social media data analytics

[@RESEARCHHUB]

Thanks for the feedback. We will do soon.

[@chinomnsodaniel7234]

Sir, please can I reach you directly.

[@simoneveggio4867]

Dear researcher, what about values where denominator is null?!?!? For example, what if my 1st period real demand is null, thus forcing S1 equal to 0? What value do I have to assume as initializing L13 (as #DIV/0! unfortunately does not help my forecasting process) ? What if my 12th period demand is null, thus my S12 is equal to 0, thus my T13 is equal to D(13) - #DIV/0! ? I've forced a IFERROR where a #DIV/0! will be substituted automatically by a 0, but I'm not that sure is the only and whenever valid solution. Thanks

[@sagarikarajapaksha-bj3l]

Thank you