Predicting the future with Fourier transforms - page 22
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
According to my observations probability of correct forecast is directly proportional to the last bar of training window (the further you start the forecast, the higher probability, the more valuable forecast)
neoclassic, you type up the statistics and post the result here. Then we'll discuss it.
And what is the meaning of this:
"the probability of a correct prediction is directly proportional to the last bar (the further you start the prediction, the higher the probability) "?
I got interested in Fourier Transform Pronosciences, and I just found an excellent Extrapolator indicator - https://www.mql5.com/ru/code/8608.
I have improved it a bit and got the following tool:
I should run the indicator, apply the script to a chart and move/modify the channel to get an adequate forecast.
If you want to get a good forecast you should find the base frequency at which this indicator has more weight, and then by changing the range of input data you should try to get an ideal picture...
I already described the problem in one of the threads
I already described the problem in one of the threads...
I also wanted to raise a question about the indicator that started this thread...
I reworked it and tested it with the original and then entered a simple sine function in the original data and found one interesting thing as a result...look at the pictures...
As you can see the function allows you to get a continuation of the curve at phase shift... the red line is the curve of the sine function the yellow
is the transformed given based on the transform function with prediction ...the predicted section has a small offset which is specially added to distinguish the predicted section with the correct selection of the input starting point and the output section we will get the predicted section perfect with the given ...here we show two different frequencies ...
In the illustrations shown here we can already see not only a divergence in the pronose direction but also a shift relative to the main signal...
... and then a question may arise - if such a picture is observed in the case of the ideal curve then what kind of prediction we can talk about with the real data...
forte928
Can you elaborate on the last post.
forte928
Can you elaborate on the last post.
forte928
Have you tried looking at the Fourier ray from a different angle?
First, divide the minute price series into different frequencies. and then decompose each frequency separately in a Fourier series. the residual is output separately. (all harmonics are found and the residue is output as noise) and so on for each frequency.
forte928
Have you tried looking at the Fourier ray from a different angle?
First, divide the minute price series into different frequencies. and then decompose each frequency separately in a Fourier series. the residual is output separately. (all harmonics are found and the residue is output as noise) and so on for each frequency.
Search the internet for the phrases "empirical mode decomposition" and "Hilbert-Huang transform" and you will get a lot of useful information on the subject.
and give up on Fourier, it has too many disadvantages that more modern methods don't have
Please list them. And is there any literature in Russian comparing these methods?
Please list it. And is there any literature in Russian comparing these methods?
Listed two posts above (you can also add wavelets). Compared, for example, in this picture (sorry for the quality):
Generally speaking, Fourier is such a bearded method that almost everything else can be considered "more modern".
There is little information (useful) on the Internet (especially in Russian), mostly just basic, so I often have to think with my own head.
Here I have managed to decompose the quotes (just a little bit undercut) into single-mode signals, below the hilbert spectrum with the noise mode excluded. As you can see, there are only 2 components. How many of them are in Fourier, you know. But I want to warn you - end effects still have to be dealt with.