Machine learning in trading: theory, models, practice and algo-trading - page 2322
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
In my opinion, for our purposes - the article is not very good, I just chose as an illustration of the approach that combines multifractality and stochasticity.
Roughly speaking, multifractal = consisting of many fractals and spectrum is the dimensions of these basic fractals. But we can play around with the notion of "spectrum" and come up with something suitable for us - for example, a function showing the degree of difference from SB at different scales.
The scale gives a larger smaller range, the spectrum, or another method of detecting non-SSB will still show what it shows, but will not connect it to the causes of non-SSB in any way. In general, access to the control of everything and everything and the processing of this data will probably give some opportunities. But it will not get into everyone's mind)))
scale gives a larger smaller range, spectrum, or other method of detecting non-STD will still show what it will show, but will not connect it to the causes of non-STD. In general, access to the control of everything and everything and the processing of this data will probably give some opportunities. BUT in everyone's brain will not get)))
Well, the servers of the brokerage companies and ECN do not let us in) We have to think it up by ourselves.)
https://www.mql5.com/ru/forum/325441/page15#comment_20589051
i.e. it should be no problem to write and debug the bot
Good article. The approach is not well disclosed, only results and half hints as done, but the results are impressive.
Good article. The approach is not well disclosed, only the results and half a hint how it is done, but the results are impressive.
I did not particularly get into it, but it seems that through some algorithm a large number of possible extensions of the series is made, of which then selects the one that by given metrics best fits the original series. I see the problem in the ambiguity of the result of such "prediction":
1) If several metrics are given, there will be a different "prediction" for each of them. If you make one compromise metric out of several, the "prediction" will depend on its particular device.
2) The "prediction" will depend heavily on the algorithm for constructing a set of possible series extensions.
The idea of moving away from parametric models is understandable and attractive, but it is not implemented here (I hope it is clear why).
I did not particularly get into it, but apparently, through some algorithm a large number of possible extensions of the series is made, of which then selects the one that best fits the original series according to the given metrics. I see the problem in the ambiguity of the result of such "prediction":
1) If several metrics are given, there will be a different "prediction" for each of them. If you make one compromise metric out of several, the "prediction" will depend on its particular device.
2) The "prediction" will depend heavily on the algorithm for constructing a set of possible series extensions.
The idea of moving away from parametric models is understandable and attractive, but it is not implemented here (I hope it is clear why).
As I understand it, the authors do not reveal the algorithm itself too much, making maxims like:
Therefore, the GenericPred method uses two basic rules:
R1: Always endeavour to keep the value of a nonlinear measure as steady as possible during prediction(Fig. 3).
R2: The new value must be chosen from a set of potential values generated from a probability distribution.
The prediction has to be pursued one step at a time because the predicted value in the current step is needed for determining the valid range of change for the next step.
As far as I guess, at first some logistic linear component is selected, and then at each step a nonlinear component is simulated, the main criterion being the stability of some set of stochastic characteristics of the series. In general, it is vague, but the result is impressive.
In my opinion the approach is somewhat similar to the one used in the package "prophet" in R.
I see that there is interest in this topic...
As far as I remember here was an attempt to implement this algorithm on R, but the articles are no longer open, at least for me, try
I see that there is interest in this topic...
As far as I remember here was an attempt to implement this algorithm on R, but the articles are no longer open, at least for me, try
There is a wonderful site that archives almost the entire Internet.
Here are copies of the first of your articles
https://web.archive.org/web/20160701000000*/https://mechanicalforex.com/2016/03/using-r-in-trading-time-series-forecasting-using-chaos-part-1.html