Machine learning in trading: theory, models, practice and algo-trading - page 3618

 
mytarmailS #:

Well, show yourself, you'll get your fill.

Thanks for the valuable recommendation.
 

The competition is interesting, unusual. Already > 5k participants. Money withdrawal seems to be in crypto, so withdrawal in RF should not be a big problem.

Prado made some vidosiki there to help.

I found out a bit late, I'll try to figure it out in time :)

 

The dataset is very complicated and takes a long time to learn. I still don't understand the meaning of what is happening. Submits are glitchy, my model has not been ranked yet. It's somewhere in the top 100 so far.

Oh, it's in the rankings


 
Prado couldn't explain what he wanted at all. He has created a set of DAG's, it is not clear what they are made of, after transformation he gets a multiclass of 8 classes. Each class refers to the relationship of one variable to another. Whether it is its cause or effect or noise or intermediate and so on.

There are no features at all, they are created from the relation of variables to each other. There are variants of creating features through correlation, f1 score, entropy.
 
Maxim Dmitrievsky #:

The dataset is very complicated and takes a long time to learn. I still don't understand the meaning of what is happening. Submits are glitchy, my model has not been ranked yet. It's somewhere in the top 100 so far.

Oh, it's in the rankings

Nice work.!!!!!!.


Give me a link to the leaderboard.

 
mytarmailS #:

beauty!!!!!!


give me a link to leaderboard

Well, there's a link above, go to the site, you'll see.
I don't know what to do next. By accident only :)
 
Maxim Dmitrievsky #:
Well, go to the link above, go to the site, you'll see
.
I don't know what to do next. Only by accident :)

this is a normal state

 
Maxim Dmitrievsky #:

The dataset is very complex

Can you post it here along with the targeting?)

 
Evgeni Gavrilovi #:

Can you post it here along with the targeting)?

There's a gig and a half of it, it's on the site in a notebook on kolab
 
Maxim Dmitrievsky #:
There's a gig and a half of it, it's on the site in the notebook on colaba

Already found it. Is the whole notebook with all the code or one parquet file sent for verification?