Machine learning in trading: theory, models, practice and algo-trading - page 2954

 
mytarmailS #:
There is also intrees package, where you can pull rules from many village models.
Forrests, boosts...

It seems to be based on XGBoost, which is considered worse than CatBoost and LightGBM. So the lightgbm package in R is also available. The question is a good way to implement models.

 
Aleksey Vyazmikin #:

What are they based on?

On the general political situation and the recent dismantling of Yandex.

 
Forester #:

1 split? Or more than one? Fxaber allocates 3 working splits by its programme (and 3 discarded ones). Total 5 splits.
Does any MO software give such possibility - to make requested number of splits by specified feature?

And the most interesting thing is to find out in which places to make splits. Fxaber does it according to the trading results. It turns out that it is not the first splits, but at the end it is necessary to add them.
Or the first splits allocate pieces of 1-4 hours and the underlying tree is trained, trying to reach the maximum on its time interval.
I am afraid that there is no MO package that does this - we will just have to split the data and train several models.

I described three variants (the second one is described in my comment to fxsaber's post), of which the first two are closer, and the third one was given as a hypothetical one. IMHO, we need a simple model with a single tree (not an ensemble) and a minimum number of features. After a new cut point is determined, the cyclic feature is recalculated (shifted) and then the usual model is used.

 
Forester #:

1 split? Or more than one? Fxaber allocates 3 working splits by its programme (and 3 discarded ones). Total 5 splits.
Does any MO software give such possibility - to make requested number of splits by specified feature?

And the most interesting thing is to find out in which places to make splits. Fxaber does it according to the trading results. It turns out not the first splits, but at the end it is necessary to add.
Or the first splits allocate pieces of 1-4 hours and the underlying tree is trained, trying to reach the maximum on its time interval.
I'm afraid that there is no MO package that does this - we will have to just split the data and train several models.

Saber knows exactly what it is trading. You can even see it with your eyes. Through all sorts of splitting you can go through endlessly. So there is no point in getting into the innards of models almost ever :) You have to think within the framework of what the classifier does. That's how to use it. It will immediately release a lot of blocked CI energy.
 
Aleksey Vyazmikin #:

Can you demonstrate a script with this functionality?

In each package there is a help for each function with description and example
 
Aleksey Nikolayev #:

It seems to be based on XGBoost, which is considered worse than CatBoost and LightGBM. So the lightgbm package in R is also available. The question is a good way to implement models.

There are all boosts except cat, as far as I remember, see help
 
Aleksey Nikolayev #:

On the general political situation and the recent dismantling of Yandex.

CatBoost has open code, its own community, and implementation in many commercial projects, so support will not stop immediately, even if Yandex will be closed.

And about the separation - even if they take out intangible assets, I do not think that in the current realities of international law will be of interest to residents of the Russian Federation, including at the legislative level.

 
mytarmailS #:
Every package has a reference for each function with a description and an example

I could say this about any language, and that people posting their codes are idiots!

 
Aleksey Vyazmikin #:

CatBoost has open code, its own community, and implementation in many commercial projects, so support will not stop immediately, even if Yandex is shut down.

Well, maybe. Though I believe in microsoft more)

Aleksey Vyazmikin #:

As for the separation - even if they take out intangible assets, I do not think that in the current realities international law will be of interest to Russian residents, including at the legislative level.

Yandex is not a resident of the Russian Federation and has never been. Now it has also physically moved and seems to have ceased to be the main search engine in Russia. It will be surprising if it is not swallowed up by google or microsoft in the next few years.

 
Aleksey Nikolayev #:

It seems that Yandex is not a resident of the Russian Federation and never has been. Now it has also physically moved and seems to have ceased to be the main search engine in Russia. It will be surprising if it is not swallowed up by google or microsoft in the next few years.

Yandex has a lot of LLCs in RF through which various projects are run. How they legally get in touch with the parent company - we still need to study it.