Machine learning in trading: theory, models, practice and algo-trading - page 49

[mytarmailS]

So how do you look for proximity through spectral components?

[СанСаныч Фоменко]

mytarmailS:
So how do you look for proximity through spectral components?

You can't.

Strictly speaking, spectral analysis is VERY inapplicable to financial time series. You see, VERY much. Because it requires stationary data, which financial time series are not.

There are examples of some particular successful (as it seems) solutions. There was Vadim Junko on the site, and he seemed to have managed to do something like that.

[СанСаныч Фоменко]

Review from here.

More and more data scientists prefer R

The results of their third annual Quantitative Business Professionals survey have been published.

Do you prefer to use: SAS, R, or Python?

Open source tools dominate overall. SAS (paid) succeeded with professionals with more than 16 years of experience, while those with less than 5 years of experience preferred R. R was also the dominant choice of analytics professionals with a Ph.

More charts are given for the two uses:

SAR.

Data from the revolutionanalytics website. It is owned by Microsoft, which not only supports the free part of the R system, but also develops paid tools.

[mytarmailS]

SanSanych Fomenko:

No.

1 ) Strictly speaking, spectral analysis is VERY inapplicable to financial time series. You see, VERY much. Because it requires stationary data, which financial time series are not.

2 ) There are examples of some particular successful (as it seems) solutions. On this site was Vadim Junko, so he seems to have succeeded in something like this.

1)ANY function can be decomposed into a Fourier harmonic series, you know ANY.....

ANY function has essentially only three arrays with parameters that fully describe ANY function and are the most objective compared to other measures - amplitude, phase, frequency.....

No offense, but if you do not understand the question and it is not necessary to get into it in a role of the teacher, here the role of the student is suitable but in any way, not the teacher ... once again no offense said

2) All I know who managed to successfully make market forecasts with neural networks, all of them used Fourier pre-processing of predictors or approximation of prices

=================================

The question is still relevant....

[mytarmailS]

Here's even this no-nonsense kid talking, watch from minute 10

https://www.youtube.com/watch?v=KUdWTnyeBxo&list=PLDCR37g8W9nFO5bPnL91WF28V5L9F-lJL&index=3

[СанСаныч Фоменко]

mytarmailS:

1)ANY function can be decomposed into a series of Fourier harmonics, see ANY.....

ANY function has essentially only three arrays with parameters that fully describe ANY function and are the most objective compared to other measures - amplitude, phase, frequency.....

No offense, but if you do not understand the question and it is not necessary to get into it in a role of the teacher, here the role of the student is suitable but in any way, not the teacher ... once again no offense said

2) All I know who managed to successfully make market forecasts with neural networks, all of them used Fourier pre-processing of predictors or approximation of prices

=================================

The question is still relevant....

You do not understand the meaning of the term "stationarity".

Good luck.

[mytarmailS]

SanSanych Fomenko:

You do not understand the meaning of the term "stationarity.

Good luck.

OMG....

[Alexey Burnakov]

mytarmailS:

1)ANY function can be decomposed into a series of Fourier harmonics, see ANY.....

ANY function has essentially only three arrays with parameters that fully describe ANY function and are the most objective compared to other measures - amplitude, phase, frequency.....

No offense, but if you do not understand the question and it is not necessary to get into it in a role of the teacher, here the role of the student is suitable but in any way, not the teacher ... once again no offense said

2) All I know who managed to successfully make market forecasts with neural networks, all of them used Fourier pre-processing of predictors or approximation of prices

=================================

The question is still relevant....

We are already tired of your claims. Not only that themselves do not put anything practical, but still require others to explain. There is no free stuff here!

And SS is right. You can apply the method, even head-on, and you will get a figure. But it won't work out of sample.

[Yury Reshetov]

mytarmailS:

1)ANY function can be decomposed into a series of Fourier harmonics, you see.....

You can do any function, if you feel like it (you can cut out your tonsils through your anus, if nobody objects), but only periodic functions are exactly reconstructed back from the decomposition. That is, although non-periodic functions can be decomposed into a Fourier series, but they are obviously incorrect because they cannot be exactly restored at the edges of a period, and the maximum accuracy will be only in the middle of a period. The period edges will always converge to the value of the zero harmonic when reconstructed backwards.

[mytarmailS]

The question was one: is it possible to measure the similarities between the functions through amplitude, phase and frequency, if so how is this done?

THAT'S IT!!! I am not interested in anything else ...

All the rest written about Fourier is a consequence of CC's answer and has nothing to do with my question