Zigzag indicator and neural nets - page 7

 
I see its greatest utility when used as a breakdown of the quote stream into classes (which the NS should recognise)
Golden words. :))
There's just one catch. These classes have to be assembled afterwards. And here you'll move away from just a neural net and get closer to AI.
One more dark forest. All you have to do is study math and biology. But God forbid you to read the trader's literature.
You'll be stuck in a jam. However it won't hurt to read Safonov.
 
Alex-Bugalter писал (а):
However, Safonov would not hurt to read it.

If you could be a little more specific about Safonov. If I have enough time, I will try to read it. And decipher the acronym AI.

 
Prival:

...


I left you a post on the last page. Have you seen it?
 
Yurixx:
I left you a post on the last page. Did you see it?
Yes I did, sorry I didn't reply right away. Unfortunately, there are no Stratonovich's works in electronic form. There are several pages from Tikhonov's works (the ones I work with now) and other books related to this subject. Contact via Skype. All will pass through him, here attachment does not poluchaetsya because of the volume. There are very interesting books.
Look for privalov-sv
 
If you could be a little more specific about Safonov. If I have enough time, I will try to read it. And decipher the acronym AI.
  • V.S. Safonov "Trading an additional dimension of decision-making" (Search on the spider, there was one)
  • AI - Artificial Intelligence, AI, artificial intelligence.
 
Prival:
Unfortunately, there are no Stratonovich's works in electronic form. There are several pages from Tikhonov's works (the ones I work with now) and other books on the subject. Contact via Skype. All will pass through him, here is not poluchaetsya because of the volume. There are very interesting books.
Look for privalov-sv


I do not have Skype, but okay. "Principles of Adaptive Reception" would like to see, of course, but I've already downloaded it for a couple or three months of work. Individual pages don't make sense. I am not a specialist in these matters, I need it from the beginning. :-)

Just post the titles of Tikhonov's books that are relevant, just like you did with Stratonovich. Tikhonov is too common a surname, you can't find it that easily.

 

Yurixx

Try to find this book. Tikhonov V.I., Kharisov V.N. Statistical Analysis and Synthesis of Radio Engineering Devices and Systems. -M.: Radio and Communications, 1991.

It is a book for universities, it has examples which help to understand this complicated mathematics and the presentation is quite consecutive. According to my information the reprint of this book is planned, but it will probably be at the end of next year.

 
Second edition, revised - 2004
 
OK. Thank you.
 
Prival:

to Piligrimm

Zig-Zag is a good indicator that has an interesting property. It's just that everyone sees its use differently. Using it in its pure form on the NS input is useless IMHO. You confirm that too, that you have had to refine Zig-Zag.

I see its greatest usefulness when used as a breakdown of the quote stream into classes (which the NS has to recognise)

Piligrimm could you please clarify, if you don't mind.

.. .the direction of the corresponding trend... What is the output in your NS (what do you mean by trend ?).

...% output of modelling and prediction error... What do you have as a model, and for what time interval is it possible to make a prediction.

You can use Zig-Zag in its pure form, in the beginning I did, I did not argue that it is useless, and my post wanted to disprove statements like that heard in many forum threads, although the prediction accuracy using standard Zig-Zag is lower than in my last indicator, but I altered Zig-Zag for additional accounting trend dynamics, A new break point in my indicator is formed not only when the trend (read trend) direction changes in accordance with the specified threshold, but also when the trend dynamics changes, in this case the trend changes its slope angle without changing its direction, which is a very informative indicator and account dynamics essentially increases the accuracy of forecasts.

The indicator data is used as an output signal relative to which the network is trained.

The % error is a result of running the network on a test sample that does not overlap with the training data sample.

I do not consider what the input indicator shows as a trend, but only a trend model, as a trend is inherently much more complex and a model is the result of synthesis of an artificial trend reflecting the current trend only with a certain degree of accuracy, due to many limitations imposed during modeling. A trend model is also what we get after the calculation of the produced trend, but in this case the trend model reflects not only the history, but also shows the direction of the trend.

The forecast is not bound to the time intervals, each time a new bar arrives on М1 it recalculates the whole expert system taking into account the new data, and either it confirms the trend before or changes the direction of the forecast when the trend reverses.

SK - the purpose of my testing was to check the indicator in dynamic mode, but not to try to get the maximum profit with its use, so I did not try to improve the Expert Advisor. The result is very good from the point of view of checking the dynamic characteristics of the indicator but from my point of view it is mediocre when I use it together with the Expert Advisor, the drawdown is too large, I would not use such an Expert Advisor for real trading.

But I am absolutely sure in one thing, this indicator will work equally in any market, no matter what timeframe I would check. It follows from the principle of its construction and is also confirmed by my numerous observations of its work during the year on a demo account.

I don't want to spend resources on downloading large quotes history for the sake of checking something that I do not doubt, moreover I am not planning to use this indicator separately, but together with an expert system that will make forecasts - the results will be incommensurable with what I have now.

Reason: