What are the data prediction methods?

Question

higher mathematics

Hello everyone, I have a small problem. I need to predict the graphs in my program, for which I need the appropriate mathematical formulas.
I found only these formulas, but I can’t figure them out in any way.

Yt = Tt + St + et Y is the value of the quantity; T - trend; S - seasonal changes; e is noise.

Mt = 1/12*(Yt + Yt-1...+Yt-11)

Ln(Yt+1) = 12 × Mt+1 – Ln(Yt) – Ln(Yt-1) – ... – Ln (Yt-10)

Yt+1 = exp(Ln(Yt+1))

If it's not difficult, please explain to me how to calculate it. Preferably with an example.

Link to source:

https://www.e-xecutive.ru/wiki/index.php?title=Pro...

PS All kinds of neural networks do not work, because you will have to predict a lot and at once.

Answer 1

It is not clear, you are citing an article, essentially an elementary introduction to the problem, where everything is simplified to the limit and explained in detail. What else can we explain to you further? And what does "how to count" mean, if you even wrote the formula yourself?
Well, except to say that people have been dealing with the problem of time series forecasting for about 70 years, have come up with a bunch of methods, algorithms, approaches, devote a lot of time and effort to studying this issue, write books. Universities study such a course, sometimes for several semesters. And you want "not difficult" and with examples. And there are a dozen libraries and software packages, from python scikit-learn to SPSS, each with a couple of dozen forecasting methods. But in order to deal with them, yes, you need to strain and figure it out.
In addition, a completely mysterious phrase about the need to predict "a lot and at once" as the reason for the "non-channeling" of neural networks. I am not a fan of them, and I think that the method should always be selected for the task, but what does "a lot and at once" have to do with it? And what about neural networks, do I work a little and once a week? However, if you don’t want a neural network, use other prediction methods, which, as I said, are many and different.
You can get acquainted with them, for example, here:
https://otexts.com/fpp2/
https://habr.com/ru/company/ods/blog/327242/
https://www.yakaboo.ua/book- at-prakticheskij-anali...
statsoft.ru/home/textbook/modules/sttimser.html
https://www.wiley.com/en-us/Machine+Learning+for+T...
https://habr .com/ru/post/180409/
https://www.youtube.com/watch?v=Rmh6b96u6UU
https://www.springer.com/gp/book/9781475777505
https://www.twirpx.com/file/560509/elib.cspu.ru/
_ xmlui/handle/123456789/4811?show=full
etc. to infinity.
In general, as specific questions appear - come back, we will try to figure it out together. Only with specific questions, and not with the question "but explain to me EVERYTHING, with examples and how to count it."