Machine learning in trading: theory, models, practice and algo-trading - page 1217
![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
the equivalent in R is Caret http://topepo.github.io/caret/index.html
essentially the same thing, universal libs.
Uh-huh. The docks are in typical R style. Like R itself, written especially for masochists.
The most promising ZZ is as follows.
The point is that the trend predicts only the beginning of the ZZ shoulder, and the end of the shoulder predicts the beginning of the trend reversal. That is why I took the end of the previous shoulder + the beginning of the next shoulder as a teacher. If we make a teacher ternary cutting off the middle shoulder, it is quite a decent teacher. But the problem of predictors to it is in all its glory.
If I remember correctly, you're using point ZZ, not bar ZZ, right?
I don't quite understand the terminology - is the leverage a segment of ZZ? If so, then " end of previous leverage + start of next leverage" is the index in pips that should have been predicted, but I didn't understand at what point the prediction should take place.
Did you make predictors by ZZ, or did you only use other predictors?
Lying on the couch, studying https://scikit-learn.org/stable/user_guide.html. This is great. ))
For those who have not yet read it, I strongly recommend it.
Zy SanSanychev R rests, and nervously smokes in the hallway. It is pathetic. ((
I don't want to be an advertiser like sanych, but I'll tell the truth, R is the best language for research, only a pathetic theorist who never touched it would argue with this....
the equivalent in R is Caret http://topepo.github.io/caret/index.html
essentially the same thing, universal libs
Yes, Max is right, it's analog, and you know whatYuriy Asaulenko? It's just one R lyric, and there are thousands of lyrics for all music styles and tastes.
For example you can
download quotes through one lib
Make the spectrum analysis through a second library.
perform technical analysis via a third one
Do data mining in the fourth one
do text mining of tweets or Facebook or a site or any text, with the fifth
combine it all into a model (neuronka, scaffolding, boosts, svm, nmm, take your pick) and optimize, crosvalidate via the sixth library, let the same Caret
and test in the strategy tester in the seventh lib
+ the richest and most advanced visualization
And all this in one language in one code, and all this fit into 50-100 lines of readable code....
What the hellscikit-learn, wake upYuriy Asaulenko, you have no idea what you're talking about.
I don't want to be a publicist like sanych but I will tell the truth, R is the best language for research, only a pathetic theorist who has never touched it would argue with that....
Yes, Max is right, it's an analogue, and you know whatYuriy Asaulenko? It's just one R lib, and there are thousands of libs for all trends and tastes.
For example you can
download quotes through one lib
Make the spectrum analysis through a second library.
perform technical analysis via a third one
Do data mining in the fourth one
do text mining of tweets or Facebook or a web site or any text, with the fifth
combine it all into a model (neuronka, scaffolding, boosts, svm, nmm, take your pick) and optimize, crosvalidate via the sixth library, let the same Caret
and test in the strategy tester in the seventh lib
+ the richest and most advanced visualization
And all this in one language in one code, and all this fit into 50-100 lines of readable code....
What the fuck isscikit-learn, wake upYuriy Asaulenko, you have no idea what you're talking about.
python plus you can write a full-fledged program with a gui, a website, an application... and not in R :) + Python is slightly faster than R
plus you can write a full-fledged program with gui, a site, an application... and you can't in R :)
You can't, but R is easy to integrate into any language and python is no problem as far as I know.
python is slightly faster than r
Yes it is.
But I wrote in highlighted that r-ka is a tool for research, not for production.
Is it possible in python to do all what I listed in one code using all ready made? I doubt it....
Yes you can, but r-ka can be easily integrated into any language, there is no problem with python, as far as I know
It's faster.
But I wrote in highlighted that r-ka is a tool for research, not for production.
Is it possible to do all I listed in python in one code using all ready made? I doubt it....
and even more )) just like the centralized repository of packages https://pypi.org/
or a github install
Google IPython, tooand even more )) just like the centralized repository of packages https://pypi.org/
or install from github
Google IPython yetwell then great
The tools are there, all that's left is to build a house))Gentlemen!
I remember, Alyosha, even before he was killed by investors, argued that the first thing to check his neural networks and forests on a simple Gaussian random walk (but not on a sine wave, as suggested by the reader Asaulenko). Moreover, he said that in this case, the prediction accuracy was 100%. And then he transferred his system to real BP.
Has anyone else here conducted similar experiments? What are the results?
Once again, if you know how to predict any of the processes (with or without memory) - just separate it from the real VR.
divide et impera
everything is perfectly predictable, some simple or compound f-i... but the market is not a simple f-i :)
example https://www.mql5.com/ru/forum/285454/page3#comment_9323441
Uh-huh.
Can I see the scoreboard?
Max, if you don't have time - give this task to an Indian. Well, it's unrealistic to do everything yourself - I understand that very well.
I gave you the link above...
an indian can't even figure out the matrices, unfortunately ))
Okay.
I understand that the main problem here is trying to work with the whole BP, which is very complex indeed.
Has anyone here tried to extract pieces (clusters) from BP according to any characteristics? Which ones? Do you have results in tabular form?
Please understand my impatience - the Grail is needed like air.
I was dividing, for example, by trading time - trade only low-volatility night time, for example. Yes, the results are improving a bit.
+ I'm experimenting to recognize current n-bar distribution of quotes and accordingly apply different strategies to them. But everything is in pile, there are no tables :)
At the moment I'm monitoring one of the versions of bot, the results are not so good this week, now I'm just comparing deals with tester and real - everything coincides.
This is how it should ideally be, on the left of the red line trade with new data, on the right - trade with history, which was training
The training is purely on returnees with different lags, during the training several best returnees are chosen
This version is from my last article, I don't remember if I changed something or not.