Machine learning in trading: theory, models, practice and algo-trading - page 1440

 
Ivan Negreshniy:

I've noticed this too and applied it, but as far as I understand it, it works because the data is crap, and this trick helps to train on a range of outliers. If I don't do it then model fits one broker or even stops working sometimes after some data reloads on the same terminal.

In short, I wrote what reduces the error, there are other more sophisticated tricks like crossvalidation but in a different manner and so on. All of these together make a pretty good model. If it's about finding patterns, it's entropy/mutual transformation/pre-processing methods.

Everybody's got their tricks there.

 
Maxim Dmitrievsky:

I have solved these problems, if not completely, then at least something works, errors on the feedback 0.1-0.2 (classification), entropy is worse but it is understandable, this metric is responsible for other things

of course... but in fact:

Shame on Maksim Denisenko

 

The most hilarious thing is that this scoundrel dares to sell this shield in the market.

I'm going to remind you what it is everywhere this scoundrel writes:

shame logo

There is nothing but this:

Total disgrace for Maksim Denisenko

 
Kesha Rutov:

Of course, of course... but in fact:


Well, if you are interested in discussing a test demo, it would be more logical to ask some leading questions, why so and not otherwise, what is the root of the evil, etc.

i.e. a constructive dialogue. Otherwise, this quacking of various lunatic magicians is not interesting to me. You can continue to give in to your illusions.

See that you are not banned for advertising, by the way. I do not advertise anything to anyone, it's my own business.

Everything that is discussed is in the public domain in the articles.

In this regard, you - a zero without a stick. No articles, no useful information, no monitornig. Nothing.

But I do not expect adequacy from such personalities, too low level.
 

shame logo

Here's the right logo for the product of fraudster Maxim Denisenko

 
Kesha Rutov:

here's the right logo for the product of fraudster Maksim Denisenko

cps )

 
Maxim Dmitrievsky:

oops )

remove it from the marketplace, at least save the remnants of honor

 
Kesha Rutov:

remove it from the marketplace, save what little honor you have left.

I'm flattered by your attention to my work, but no

 
Maxim Dmitrievsky:

If it is a matter of searching for patterns, these are entropy/mutual transformation/pre-processing methods

how to identify the alphabet or the amount of information in the CD?

And again to the trivial question - what to use as an input? - closing price, the entire OHLC bar, the alternation of bars? what is the TF....

Last year I spent a lot of time reading this forum (search for entropy, topics from 6-7 years ago)

and recovering knowledge from university - I was taught information theory, it's an interesting discipline, but to apply it you need to know what input data we have - if you don't have such knowledge, it's cryptography, then, as in our case (CR), everything boils down to bruteforce - that's what we do in IR))

 
Igor Makanu:

how to identify the alphabet or the amount of information in the CA?

And again to the trivial question - what to use as input data? - closing price, the whole bar OHLC , alternation of bars? what is the TF....

Last year I spent a lot of time reading this forum (search for entropy, topics from 6-7 years ago)

and recovering knowledge from university - I learned information theory, interesting discipline, but to apply it you need to know what is the input data - if you don't have such knowledge, it's cryptography, then, as in our case (CR), everything comes down to bruteforce - what we do in IR )))

Exactly on the 1st question, there is an articlehttps://habr.com/ru/post/127394/

On the basis of this you can already make some assumptions

about brutforce - still have to search in a limited area, otherwise it can go on forever.

What I do myself - I do not know how to explain in 3 sentences. Just going through the options based on experience.

Regarding the article and market memory - see in person, threw recently. Only 3 people know (the most competent) so far, the topic has not yet been fully explored :)

Waiting for feedback, as they say, but also doing it myself
Теория информации в задаче проверки гипотезы о независимости значений, принимаемых случайной переменной, на примере индекса DJI
Теория информации в задаче проверки гипотезы о независимости значений, принимаемых случайной переменной, на примере индекса DJI
  • habr.com
Попробуем проверить гипотезу о том, являются ли приращения значений индекса DJI статистически независимыми. При этом в качестве референсного источника данных, с которым будем проводить сравнение, возьмем искусственный временной ряд, сгенерированный из собственно приращений исходного ряда, но при этом случайно перемешанных. В качестве меры...