Machine learning in trading: theory, models, practice and algo-trading - page 2117

 
Maxim Dmitrievsky:

This is a serious study on class balancing, I haven't finished yet

Yes, there are a lot of methods - now we need to figure out which ones are effective.

Here's a compilation of various methods description.

Алгоритмы сэмплирования — Викиконспекты
Алгоритмы сэмплирования — Викиконспекты
  • neerc.ifmo.ru
Сэмплирование (англ. data sampling ) — метод корректировки обучающей выборки с целью балансировки распределения классов в исходном наборе данных. Нужно отличать этот метод от сэмплирования в активном обучении для отбора кандидатов и от сэмплирования в статистике [1] для создания подвыборки с сохранением распределения классов. Неравномерное...
 
Maxim Dmitrievsky:

I don't know what you're discussing here.

there are two feature spaces (took 5 main components of each)

In the case of random sampling of transactions:

In the case of clustering into 2 clusters:

Task: find the trade-off between the correct labels and a good partitioning of the classes.

In the case of simple clustering, the labels are of course unsuitable for trading

In the case of sampling trades - the feature space is not good

Something similar was already done in R with mytarmailS. I tried to teach models within classes and then combine them into a single model - the models were more overlearned. Maybe I didn't do enough testing.

 
Maxim Dmitrievsky:


Gorgeous pictures. Perhaps this is what the Grail looks like in the projection from N-dimensional space...

 
Alexander_K:

Gorgeous pictures. Perhaps this is what the Grail looks like in projection from N-dimensional space...

Write more often )) it lifts your spirits ...

Maxim Dmitrievsky:

I don't know what you're discussing here.

Since you finally got to psa, do a decomposition with non-normalized price in window of say 20, and see what you get, I think it will be interesting...

 
Aleksey Vyazmikin:

Try it and tell me about it later :)

I wish I knew English like a native ( and they are called BANANAS, google collects so much garbage on this search (

I'll have to try as soon as I get to python) I wanted something more detailed in Russian.

 
Alexander_K:

Gorgeous pictures. Perhaps this is what the Grail looks like in the projection from N-dimensional space...

You can't get enough of pictures...)
 
mytarmailS:

Since you finally got to psa, do a decomposition at the non-normalized price in a window of say 20, and see what you get, I think it will be interesting...

well, it's morning in the village...

 
Maxim Dmitrievsky:

Well, it's morning in the village...

what?