Forecasting Methods made simple - Exponential Smoothing

This is the second video on the series of Forecasting Methods made simple. Here we discuss the exponential smoothing method. We discuss the relevance to the smoothing factor and there is some insight into choosing the apt value for the smoothing factor.

[@kingmakerkrish]

Dear Piyush, your videos on forecasting methods so easy to learn and helpful..... Great. Thank you

[@kyoshisfan4053]

you're so clear when explaining it. thank u so much!

[@piyushashah1]

Thank you for this message.

[@9934max]

In most of the videos i have watched, it is not clear how there are assuming the initial forecasting. In this video, my doubts are cleared. Thanks for making this video lecture.

[@abdulrehman75189]

Thank you so much for explaining so well.

[@arisgunaryati5407]

Thank you very much. It's very clear to understand. 👍

[@debbiedickinson3633]

I really like using this forecast because it can change when you change the alpha.

[@UdayBhargav1]

This Video Explains it all. So clear compared to all the videos out there. Im a university student and i feel this video explained me better than my professor did. Superb!

[@piyushashah1]

Thanks for dropping the comment. Extremely happy that the videos helped you learn better. Do share the videos with your friends and also have a look at other videos on my channel.

[@Mawisky]

thank youuuuuuuuuuuuuuuuu!!! seriously, thanks so much

[@gjergjbengu]

Hello Piyush. Your videos of forecasting are really nice. I have a question related to forecasting credibility. I have a case of Resource Based Method of Forecasting. (Project Rev= Resources Expected Hours x Rate and i have like 125 projects to forecast). What would you recommend as the best method to apply some kind of accuracy or uncertainty or confidence level to this case? My forecasting period extends only to one month ahead.

[@pratibhapal1]

Really simple explanation .. can you pls put a video on winter Holt's method as well

[@flowarina.ddevassy8230]

so thankful....

[@ellemason2055]

I know this is an old video. But thank you! Thank you so much!

[@swamypamu26]

I saw your all videos. The way you explained is awesome! superb examples. you made my day.Good luck sir.

[@piyushashah1]

Thanks a lot for your comments Swamy.

[@KunalKumar-id2zz]

very nice for understanding alfa ..........

[@dengtiu5216]

Amazing

[@EvilAbD]

Dear Piyush, thanks for the videos, very useful, I have a question, can I use this method for seasonal data? Why not? what if I used? how the results will affect the forcasting?

[@rohitkumbehera]

Have you done anything for the Croston's method to forecast sporadic demand? If yes, kindly share your spreadsheet template.

[@NazrulIslamshariatpur]

What is the optimum level of alpha? For sales forecast of a restaurant what should be the alpha value I may use sir?

[@peterheine6240]

Hello, I have a question about forecasting. Should plan corrections be determined during forecasting? Or is it not necessary to determine them? Why is there an extrapolation? An extrapolation is absolutely necessary.

[@chetanmundhe7899]

very good video

[@brianescobar7500]

Thank you, from failing my assignment

[@Mr234187]

please share the link for rest of the videos

[@georgemichal8828]

is this video regarding to calculating alpha using simple exponential smoothing ?

[@RaviKumar-mk4mj]

Hi Piyush, I need help with SKU level forecasting. Do u have any videos, files or links you can share with me, highly appreciated. Thanks

[@kamalgautam3956]

nice

[@mikedizon3825]

Hi Piyush - just curious, how come in other tutorials I've seen about exponential smoothing a damping factor (1-alpha) is used in the formula. It does not seem you used the damping factor for this. Is a there a reason why you didn't use the damping factor?

[@piyushashah1]

Hi Mike, it is exactly the same formula and the terms are rearranged. Try opening the parenthesis in my formula and rearranging and you will have the same (1-alpha) based formula. If it does not work, let me know and I will try to help you.

[@DailyWMO]

Can we use this method to forecast the safety incidents in the future ?

[@piyushashah1]

Generally no. Since accidents are generally dependent on some factor, poisson regression based methods work better.

[@subhashreerv]

it's easy to understand. if v want to predict share price using 5 years data how it ll be used can you please explain.

[@piyushashah1]

+Subhashree Prasad: These methods have an assumption that the past pattern will repeat in the future.I don't think this applies to financial markets and hence they cannot be used. I am assuming that there would be different methods to be used for yearly, daily and hourly predictions.

[@rajatbhatheja356]

This method applies only to stationary series. Stock market is not a stationary series. Stock market follows random walk pattern.

[@samkab677]

Hi thank u.but my question ..what does alpha represent ? And how to choose the suitable alpha ?

[@piyushashah1]

Forecast your past data with different values of alpha. Select that value of alpha with the least error in predicting the past. Look here for ways to measure the forecasting error. kzbin.info/www/bejne/jYqlfGekbJqelZo

[@muhamadfazrral3424]

NICE

[@piyushashah1]

Muhamad Fazrral thanks. Have a look at all the videos in this series and let me know your views

[@pushkalupadhyay]

If x axis show date when how to use this formula

[@ghoomketu.forever]

How do you decide the value for Alpha when using this method in reality

[@TheRohit9463]

You can use the solver option in excel which gives the most optimum value for smoothing

[@pachodomi]

Don't listen to this idiot, he doesn't know shit.

[@TWISTERZED]

You can take any value from 0 to 1

[@Yan-wj6kz]

Indonesia☝️😉

[@DailyWMO]

Would you share me the theory or articles of this method you using for this ?

[@piyushashah1]

You can find details of this method in any text book on Operations Management.

[@1rasha]

What are we forecasting ? When we already have data ??

[@piyushashah1]

Please see the answer to your previous question. Once you know that a certain alpha is good for a certain data, we can use the method shown to forecast for the next month.

[@npurohit1989]

Using these methods how can I forecast for next 3 or 6 months?

[@piyushashah1]

Exponential Smoothing and Moving averages can only be used to forecast the immediate next period. The idea is that the 'noise' or random variations in demand get smoothened by these methods. Thus as an assumption we are forced to believe that the predicted demand will be constant for all the periods in the future. If you use trend lines you can more easily forecast the future. Two things, One - the more you extrapolate the future, the more will be the inaccuracy of any model. Two - Choose the right method according to the type of variation you have in the concerned product.

[@npurohit1989]

Thanks Piyush

[@borisgotov9838]

would you give some example from the real world that really work, but with critical minimum of tests. 100+ for example that give the percent of accuracy. For example, if we have $/€ price for last 40 days and try to predict next business day.

[@ahsanahmed6412]

If two forecast given in a data how can we solve it?? Please someone tell me

[@piyushashah1]

Can you explain your question in detail please?

[@ahsanahmed6412]

2 forecast means 2 answer.. I understand from my teacher..

[@vickypatil5253]

Can you please provide source of formula

[@piyushashah1]

All supply chain and forecasting books should have the exact same formula.

[@sumukhhegde6677]

How to do this using goal seek to optimize alpha

[@piyushashah1]

I don't think so, since we can't have a specific goal for alpha. However, you can definitely use Excel Solver and select the alpha value that minimizes the error.

[@sumukhhegde6677]

@@piyushashah1 I've never tried excel solver before.. anyways thanks for the reply brother 😁😁

[@piyushashah1]

@@sumukhhegde6677 so guess this is your chance to learn it

[@borisgotov9838]

Dude, you must rename this video to post-cast :D

[@piyushashah1]

All these quantitative methods need past data to assess their suitability. And that is why these methods are often criticized as 'driving forward by looking at the rear view mirror'.Yet, the methods are about driving forward - and hence 'fore'cast. Let me know if you need more help on this topic and learn more.

[@borisgotov9838]

@@piyushashah1 The column F shows the past, not the future. Why you call this method forecast?

[@usmanshah739]

Is alpha = margin ???

[@piyushashah1]

Alpha is the degree of forecast sensitivity to past demand. So, if alpha is zero, forecast will not be affected at all by previous demand. And, if alpha is one, forecast for next month will equal to demand of previous month. In the video, as the alpha increases, the forecasts tend to get smoother. Check the graph displayed for this .

[@usmanshah739]

Piyush Shah thank you.. how to calculate degree of forecast sensitivity.? Please advise

[@piyushashah1]

Am not sure if that can be calculated. The idea is based on the forecasters belief of how much weight to assign to more immediate periods. If she wants to assign higher weight to immediate periods, the sensitivity needs to be higher for the immediate periods and the alpha thus has to be higher.

[@saurabhsingh1981]

Where did u forecast?? Against history u generated forecast, any one can do that

[@piyushashah1]

Well, time series method is about using historical data to forecast, and there is no doubt that with the correct formula, anyone can do it. Exponential smoothing method uses the data of actual and forecast of the previous month to generate the forecast of the current month - thats just the way it is.

[@lukaivezic3924]

@@piyushashah1 And what if you need to do forecast for whole next year, from this moment? But suppose you have only initial actual demand. What would be your actual demand for April, for example?

[@piyushashah1]

@@lukaivezic3924 Typically, more past data leads to better future forecasts. If you have one period data, these methods will then say that all future forecast are same as the one data point. So, if you just have one period past data, these methods may not work.

[@pachodomi]

What the fuck? This makes absolutely no sense.

[@ganeshpamarthi3438]

hlo sir can i contact you sir for helping my project

[@piyushashah1]

Post it here