Predicting the future with Fourier transforms - page 43

New comment
 
Integer:


Isn't transformation a series expansion?

Decompose, add, get the same thing, works on anything.


When applied to financial markets, it should be a chain - decompose, add, get not the same, but profit. Otherwise all this fiddling makes no sense ))))
 
LeoV:

Agreed, but I thought we were talking about making a profit here....((((


You decompose it, tweak the harmonics, add it up - it's a filter, an advanced MA with infinite possibilities for adjusting it.

Approximation is also a method. You approximate it, look where it is directed.

 
Integer:


Isn't transformation a series expansion?

Decompose, add, get the same thing, works on anything.

No. A series is a series. A F. transformation is a generalization of F. series to a wider class of functions. That is if historically. From the theoretical point of view, the Fourier series is a special case of the Fourier transform when the function is periodic.
 
Integer:


You decompose it, tweak the harmonics, add it up - it's a filter, an advanced MA with infinite possibilities for its regulation.

Approximation is also a way of doing things. You approximate it, look where it is pointing.


Well, where is the profit? ))))
 
LeoV:

OK, where's the profit? ))))
Guess where it's headed - profit.... guessed wrong - moose((.
 
Integer: Have approximated it, see where it is going.

alsu:Guess where it's pointing - profit.... didn't guess - moose((.

This is where the question arises - is it possible to make money in the financial markets this way? Since "watching the direction" is a prediction on every bar, something we talked about above and you said it doesn't work.

Judging by some evidence, on papers - apparently possible, due to large trends. On forex - apparently not, as the trends are not big.

 

The sunspot cycle is approximately 11 years. They have found correlations between this cycle and epidemics, revolutions etc. I always thought it was something close to a sine wave.

And this is what the spot spectrum actually looks like:

If converted correctly from dB to times, the cycle frequency in amplitude is only 3 times the other components.

So it is quite applicable to the market.

 
alsu:
No. A series is a series. A F. transformation is a generalisation of F. series to a wider class of functions. That is if historically. From the theoretical point of view, the Fourier series is a special case of the Fourier transform when the function is periodic.

The Fourier series is the sum of sines and cosines of different periodicity and amplitude, and any curvature can be drawn with this series.
 
Integer:

A Fourier series is the sum of sines and cosines of different periodicities and amplitudes, any curvature can be drawn with this series.
Not quite. If there are a finite number of sines and cosines, it's a series. But this number is finite not for all, but only for periodic functions. For all other functions we obtain some generalisation with infinite number of sines/cosines (and infinitesimal intervals between them)
 
Integer:

The Fourier series is the sum of sines and cosines of different periodicity and amplitude, and any curvature can be drawn by this series.

That's what I mean, because if this set of sine waves changes at different intervals (or if the interval changes randomly), it's not Fourier's fault, it doesn't have to take it into account.
1...363738394041424344454647484950...55
New comment