If anyone has a problem, please finalise AdaptiveExtrapolator v1.1 - page 12
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
But I've seen what I expected to see, you won't make money with it.
Agreed
I ran geExtrapolator_sig_1 but had to comment out cwCalculated and int ForeCastExtrapolator() works without them,
I put it all into one file.
But I would like to discuss the problem: why does it lie?
It doesn't lie. It calculates harmonics and extrapolates them forward. If the extrapolated harmonics did not coincide with the future price, it means that the extrapolation window was selected incorrectly and the harmonics are irrelevant.
Now this is a relevant discussion.
I.e. if you get the window right, the modelling (prediction) quality will be higher.
What do you mean harmonics are irrelevant?
Yes, the main problem is the window selection. And harmonics are irrelevant - oscillations with such phases, amplitudes and frequencies do not exist at all, or have stopped.
I don't agree with you about the window, it's very simple, do the FFT for the maximum number of available bars and see in which area there are
the perturbations, then you pick the window. Since all methods are formalized, it only remains to realize automation of this process.
I personally see another problem. I do not know how to describe it correctly, I will show an example: Set one harmonic with a period multiple of the window
2*pi*i/Period --> this harmonic is interpolated by two harmonics and now change the period so it is not a multiple of window -->
such harmonic is interpolated by the disturbance field and if we null the field and keep the largest one, the result will not correspond to
to the original, i.e. one non-multiple harmonic is given by a dozen multiples. Imagine how many harmonics the market can have...
Here's an indicator for clarification...
Set the frequency from the range 4,8,16,32,64,128,256,512... and any other frequency and see the difference.
It doesn't lie. It calculates harmonics and extrapolates them forward. If the extrapolated harmonics do not coincide with the future price, it means that the window for the extrapolation is wrong and the harmonics are irrelevantGhf
That's right, Uncle Fyodor...
I tried the F.F.T. It's the same principle and it works...
I've tried to apply this indicator on extrapolation of another function, in some cases it matched, in some cases it turned earlier, in other cases it lied... That's why I decided to take another way... we take smoothing FFT function and additionally smooth the ends by spline interpolation... and then based on the complex processing set we form the process of upcoming history...
Can you throw me the spline interpolation formulas?
Look at the indicator I attached above to make it clearer. What do you think?
Yes, the main problem is the window selection. And harmonics are irrelevant - oscillations with such phases, amplitudes and frequencies do not exist at all, or have ceased.
I totally agree with you about the window selection, here is an example in the figure where based on a perfect sinusoidal function a window was selected in which the coincidence of the signal in the future was idyllic...
the blue line is the forecast... where the blue dots are that range of input data...
the input window itself is shown on the sine curve in the light orange line (the arrow shows the boundary of the input window)
by slightly increasing or decreasing the input data the ideal sine curve will extend exactly in the direction indicated by the red line beyond the blue points boundary...
Therefore, with proper selection of the size of the input data can be achievable ideal forecast in the continuation of 20-30 points, then again it is necessary to find that input data window at which the possible continuation of the price chart, but as seen such a problem in the absence of matching apparatus this problem looks rather difficult ... as the market is constantly present several frequencies that simultaneously give a picture of uncertainty ...
I forgot to mention the gaps... at rather large gaps of 10 points and above Pronoz no future at all realistic because contribute to the system a strong impulse response, which creates a subsequent decaying process sedation, which in turn vrivnosit the presence of a large influence of low-frequency component ...