Machine learning in trading: theory, models, practice and algo-trading - page 1253

 
Aleksey Vyazmikin:

Yes, the trawl by Doncian channel is used (as in the preparation of the sample for training). Closing only by SL here, the result can be certainly improved by using smart TP, but I still have not got enough time to finish the idea.

i can't say how much i'm right, but i usually test with fixed lot and without trailing, i see the curve going up, then i can improve it

Do you have tests without trailing and with fixed lot? i want to know how different they will be from the picture above

 
Igor Makanu:

It's hard to say how right I am, but I usually test the TS with a fixed lot and without trailing, if I see a curve, then I can improve it

Do you have tests without trailing and with fixed lot? i want to know how different they will be from the picture above

The lot is fixed. I don't know, maybe I should adjust TP and SL, but what's the point of it, if the original concept of TS was based on trawl? My TS is simple - entry at moment when vector ZZ changes, on the next bar stop loss is at the point where vector ZZ changes in the opposite direction. This system is convenient for learning and then you can improve it by shifting the entry point, taking into account the signal for entry and adjusting TP - you can really add 30% for a couple of days of adjustment. A fixed SL spoils the concept, because it violates the logic of the TS. The trawl is slow enough and its triggering (closing by SL) symbolizes the moment of taking a new decision on further trading operations.

 

In general, I've looked through a lot of theory about approximate trajectories, and a lot of math... I'll have to study it for a while. So far I see potential in terms of iterative transformation of traits (i.e. you can add a degree of interpolation at each iteration, and the trait will change), but there are some other tricks that I have not yet understood

https://arxiv.org/pdf/1603.03788.pdf here is a general use for machine learning

https://arxiv.org/pdf/1307.7244.pdf and here is piecewise linear interpolation for quotes and something else

Of course it may be relevant only if you want to make a fully self-learning TS, the high level, so to speak.
 

Gentlemen expert builders, are there any articles on how to do machine learning and neural networks, by standard means?

There are in the standard library:

  • Statistics - functions for working with various distributions from probability theory
  • Fuzzy logic - fuzzy logic library, which implements Mamdani and Sugeno fuzzy logic inference systems
  • ALGLIB - data analysis (clustering, decision trees, linear regression, neural networks), differential equation solutions, Fourier transforms, numerical integration, optimization problems, statistical analysis, and more
But without an article, you can't eat that kind of stuff.
 
BillionerClub:

Gentlemen expert builders, are there any articles on how to do machine learning and neural networks, by standard means?

There are in the standard library:

  • Statistics - functions for working with various distributions from probability theory
  • Fuzzy logic - fuzzy logic library, which implements Mamdani and Sugeno fuzzy logic inference systems
  • ALGLIB - data analysis (clustering, decision trees, linear regression, neural networks), differential equation solutions, Fourier transforms, numerical integration, optimization problems, statistical analysis, and more
But without an article, you can't eat that kind of stuff.

Lasciate ogni speranza, voi ch'entrate.

 
Yuriy Asaulenko:

Lasciate ogni speranza, voi ch'entrate.

That's it, he who goes to the Ministry of Defense will go to hell.

 
Maxim Dmitrievsky:

Learned to do preprocessing correctly, well, the classification errors in both samples <0.1, and begins to work on new data, though datasets are huge, long to learn, so I will show hours on a small story

How I did it - I won't tell in a publique :)


Greedy :( .
 
Aliaksandr Hryshyn:
Greedy :( .

did not have time to delete, noticed )) may write later, I need more tests

Just here everyone communicates through hints, they do not write anything about data preparation, everyone has his own samovar
 
Maxim Dmitrievsky:

did not have time to delete, noticed ))) may write later, more tests are needed

Just here everyone communicates through hints, do not write anything about the preparation of data, everyone has his own samovar

it's because of the principle of "tin can" opens one and eats all )))

 
Anatolii Zainchkovskii:

it's because of the "tin can" principle that one opens and everyone eats)))

The Forex bank is so big that one can't eat it all by himself))) At least because it includes all the other banks in the world