Machine learning in trading: theory, models, practice and algo-trading - page 2041

New comment
 
elibrarius:
less trees, less depth, might be enough, at least just to test
Or a better comp)
 
Maxim Dmitrievsky:
Where is the time there? The day of the week, the day of the month, the hour of the day, the minute of the hour. Time is a continuous value, and here we have ordinal categories.

Aren't these not intervals of time?

I can teach with categorical attributes as well - let's check the efficiency, which posts to make categorical - everything related to time?

 

I do this:

1) I create an array of string indexes with a length equal to the number of strings, fill it with values from 0 to N strings

2) I shuffle this array

where RandomInteger() is any variant of the RNG

3) then I take all values of these indexes in a loop and use them to get the needed string from the main array; it turns out to be a pseudo-random string after mixing the indexes

Thanks, I will try to use it!

 
mytarmailS:

Has anyone tried to do a classification for a lot of classes, say 10k?

Does it work at all?

I can try running it on CatBoost - reset the sample.

 
I tried to calculate the average column in the usual code, and was pleasantly surprised. It was calculated in a fraction of a second, it took longer to read the index file into memory (30 seconds). So the bruteforce should go quickly.
 
Aleksey Vyazmikin:

Aren't these time intervals?

I can also teach with categorical signs - let's check the effectiveness, which posts to make categorical - everything related to time?

you can also try to remove the day of the month.

 
Rorschach:

You can also try deleting the day of the month.

In and out?

 
Yes
 
Igor Makanu:

This is exactly the case, but as long as you consider atoms, and as soon as you decide to consider the crystal lattice, "it's all over" - in most cases a very difficult problem to solve, I read once, by the way they use neural networks - they get the result of the modeling process faster

it is the same with the markets, while we look at one or two ticks, there are not many variants - either up or down, everything becomes complicated if we proceed to the study "here they opened an order, and here they closed it" )))

Yes, this is a certain living structure, but it has a behavioral model - let it be called a patterns or a market context, it doesn't matter

It is possible to study the behavior on the history, the only thing left is to decide what to do in the future

Just understanding the properties of the crystal lattice was solved through understanding the arrangement of the simplest particles, not only the atom and the electron and its spins, but also the quarks and a bunch of the simplest particles. Without this understanding, there was no understanding.

Which is in the market data. Well apparently education is also there, i.e. it is necessary to analyze textbooks in schools and institutes of the analyzed countries and draw conclusions)

Predicting the behavior of society is no more difficult than predicting the fate of the individual / person. It is necessary to guess if the husband will be faithful or not.

 
Rorschach:
Yes



Got more trees, quite good - on the exam sample accuracy is more than 60%.

It turns out all the same time of finding a trade, stops and exit are intertwined, which is logical - if the trade is already long open, the stops are not knocked out, probably from the fact that they are large...

Models to attach?

1...203420352036203720382039204020412042204320442045204620472048...3641
New comment