Machine learning in trading: theory, models, practice and algo-trading - page 817

[Yuriy Asaulenko]

Maxim Dmitrievsky:

I threw it a long time ago, now I came back because there are interesting points (not even on the Bayesian NS itself, but in principle)

There is also a 2 part. Look for it on the author.

[Maxim Dmitrievsky]

Yuriy Asaulenko:

There's also a part 2. Look it up on the author.

Yes, I know.

[Andrey Dik]

Maxim Dmitrievsky:

Yes why, let's poke around the topic of probabilities, if no one has any ideas

I just feel that this is the last thing left, and if it does not improve TS, retraining can not overcome in principle

There are already some interesting ideas that I will withhold for now in order not to excite the minds of "sympathizers"

here is just an interesting article

https://habrahabr.ru/post/276355/

It's interesting. It seems like everything has been known for a long time... but it stimulated fresh thoughts.

It's probably not what's in the material that's important, but how the material is presented.

[СанСаныч Фоменко]

Alyosha:

It does not matter, mashki with martin, ML on python or R if some guard or clerk will twist knobs based on his "intuition", the result is the same, Fa at least knowingly offers a lame GARCH, whose past price is the best forecast of the future, Fa is not trying to make people hope for nothing, in this he is more honest.

For the hundredth time:

It is obligatory to start with choosing only those predictors which influence the target variable. And then all the datamining.

2. There are two models:

• Classification, which should automatically find patterns.
• GARCH which must take into account statistical characteristics of the time series

Training of models with cross validation if possible.

Evaluation of models outside the training file

5. Test run in the tester.

And for the hundredth time: ALL Stages are MUST!

Having done all of these, you can make the assumption that the depo will not sell out at once!

Let's go, men! Finish hanging out on the forum and with quiet joy to implement the outlined plan for R.

Three cheers!

[Alexander_K2]

Where is Michael with the unrestrained Grail?

[Dr. Trader]

On the islands :)

[Alexander_K2]

Alyosha:

It does not matter, mashki with martin, ML on python or R if some guard or clerk will twist rolls based on his "intuition", the result is the same, Fa at least knowingly offers a lame GARCH, which past price is the best forecast of the future, Fa is not trying to reassure people, in this he is more honest.

The only smart person here is Alyosha...

[Aleksey Vyazmikin]

Can you tell me what neural network algorithm can be used to detect the logic(neuron) of the "Calc" column?

Delta<200	Delta<350	Delta>350	ZZ_D	Calc
0	1	0	1	1
0	1	0	1	1
1	0	0	1	0
1	0	0	-1	0
1	0	0	1	0
1	0	0	-1	0
0	0	1	1	1
0	1	0	1	1
0	0	1	1	1
0	0	1	1	1
1	0	0	-1	0
1	0	0	1	0
0	1	0	1	1
0	1	0	1	1
0	1	0	1	1
0	0	1	1	1
0	1	0	1	1
0	1	0	1	1
1	0	0	-1	0
0	0	1	1	1
0	1	0	-1	0
0	1	0	1	1
1	0	0	-1	0
0	1	0	1	1
0	1	0	-1	0
0	0	1	1	1
0	1	0	1	1
1	0	0	-1	0
0	1	0	1	1
1	0	0	-1	0
1	0	0	1	0

[Maxim Dmitrievsky]

Andrey Dik:

That's interesting. It seems like everything's been known for a long time... but contributed fresh thoughts.

Maybe it's not what's in the material that's important, but how the material is presented

Yes, and RBM has long been known, but now there is a lot of new research in this area that I have not read about

but the main fun is that it can be used for pre-processing features, that's what I need

... i'm dumb, so in the dipling already used ... lol... i just figured out why :) again, everything has already been invented before us

[Dr. Trader]

Aleksey Vyazmikin:

Can you tell me what neural network algorithm can be used to reveal the logic (neuron) of the "Calc" column?

It's better to use a tree for this, this model will create such a set of rules:

if( [Delta<200] >= 0.5 )
{
  return 0;
}
else if(ZZ_D < 0)
{
  return 0;
}
else
{
  return 1;
}

I wrote the code quickly, the model gives the result either as a text or as a picture

There is a description in the article how to do it in R:
https://www.mql5.com/ru/articles/1165

In the "Model" tab, select the tree. Set "min split" and "min bucket" to 1. Create a model, and then click on the Draw button and you will see this picture. Rules - show the rules in text form