Indicators: Fourier extrapolation of price - page 2

New comment
 
Prival:

No, it's not lost. Except there's this huge rock there, and everything crashes against it. As an electronics guy, you'll understand. I'll try to keep it consistent.

That's very clear. Thank you. About the dt sampling step variation. Do you think it is random or has any predictable character? If the variation of dt is predictable, we can write a Fourier series into it. We get two nested Fourier series: the first for the prices themselves and the second for the time step dt. In this case we get something similar to a frequency (or phase) modulated signal:

harmonic h(t) = mh + ah*cos(wh*t) + bh*sin(wh*t)

time t = mt + at*cos(wt*t) + bt*sin(wt*t)

total h(t) = mh + a*cos(wh*(mt+at*cos(wt*t)+bt*sin(wt*t)) + ...

There is a whole direction in Fourier analysis called nonuniform fourier transform.

 
Saidar:
Hey, for some reason the indi won't display in mt5 under indicators.  I did compile the indi, and restarted mt5.  All the other indis work except the two extrapolation indis you posted in the codebase.
Not sure why this happens, unless you use Windows 7, where the "real" path to MQL5\Indicators is somehere in C:\Documents and Settings\User_Name\...
 
gpwr:

.. If the variation in dt is predictable...

I don't think so. Not even predictable, I'm sure it's not predictable. There's no physics to it. It's a random variable. There's some research on it here https://www.mql5.com/ru/forum/103289.

Unfortunately I have very little experience with such processes (it's very complicated). The best option is to go beat up the person who made such a clock :-) usually helped)). But this method does not work here.
 
Yes I use Windows 7, will have a look
 
Maybe this will help. Analysing non-uniform time series.
 
By the way, why not build the spectrum as a histogram?
 

HideYourRichess:
Кстати, а почему бы спектр, в виде гистограммки, сразу не строить?

Are you suggesting this for visualisation or for finding strong harmonics from this spectrum? Thanks for the attached links in the previous post. We will read them.

 
Well, yes, for visualisation and finding strong harmonics. And one more thing, maybe we should introduce a feature to remove the linear trend on the Npast plot? I.e. we take the data on the plot and remove the linear trend from it. Then these detrended data are decomposed into harmonics.
 

An extrapolation in the form of candles would look interesting, not a line (not a criticism but a suggestion. I am weak as it is).

 

The period should be tied to time, so that it would be recalculated when switching timeframes. Otherwise, the picture is not quite clear and really contradictory.

For example, in hours T = hr * 60.0 / Period().

And it is better to adjust the series in resonance with the already passed data, like quickly scan and choose a variant by minimum RMS or maximum correlation, then the forecast is more likely to coincide with the real picture.

12345678
New comment