Machine learning in trading: theory, models, practice and algo-trading - page 2395

 
Aleksey Nikolayev:

Quite logical) a philosophical question is a philosophical answer)

The thought was something like a composition of classification algorithms. For example, fuzzy classification is done by the K nearest neighbor method, but some neighbors are discarded by another algorithm with clear classification. If this makes any minimal sense, there must be articles on this topic on the web)

Here again Prado comes to our aid (Meta labeling section)

https://towardsdatascience.com/financial-machine-learning-part-1-labels-7eeed050f32e

Financial Machine Learning Part 1: Labels
Financial Machine Learning Part 1: Labels
  • Maks Ivanov
  • towardsdatascience.com
In the previous post, we’ve explored several approaches for aggregating raw data for a financial instrument to create observations called bars. In this post, we will focus on the next crucial stage of the machine learning pipeline — labeling observations. As a reminder, labels in machine learning denote the outcomes of the random variable that...
 
Maxim Dmitrievsky:

For clearly separable, SVM may sometimes be better suited, but otherwise, either boosting or NS.

I want the algorithm to not only work out of the box, but also somehow take into account our specifics. In my opinion, we have both mixing and separation of classes. Or one could say that the degree of their mixing can vary greatly in different places in the set of attributes.

 
Aleksey Nikolayev:

I want the algorithm to not only work out of the box, but also somehow take into account our specifics. In my opinion, we have a mixture of mixed and separated classes. Or we could say that the degree of mixing can be very different in different places in the set of attributes.

The simplest case is in the link above. Next, you can use your imagination :)

 
Maxim Dmitrievsky:

looped, in fact, it is no different from onTick or onTimer in EAs

Template.

Сервисы. Уже работают?
Сервисы. Уже работают?
  • 2019.12.11
  • www.mql5.com
Билд 1961. В редакторе создал новый вид программы - сервис. В OnStart прописал: Далее запустил сервис: и сразу результат: В общем работает...
 
fxsaber:

Template.

And the python program can work as a service or directly as a bot

https://www.mql5.com/ru/docs/integration/python_metatrader5/mt5symbolinfotick_py

Документация по MQL5: Интеграция / MetaTrader для Python / symbol_info_tick
Документация по MQL5: Интеграция / MetaTrader для Python / symbol_info_tick
  • www.mql5.com
symbol_info_tick - MetaTrader для Python - Интеграция - Справочник MQL5 - Справочник по языку алгоритмического/автоматического трейдинга для MetaTrader 5
 
Maxim Dmitrievsky:

The simplest case is at the link above. Then you can apply your imagination :)

Thanks, hope it helps. The Prado party is our helmsman! )

 
Maxim Dmitrievsky:

And the python program can work as a service or directly as a bot

https://www.mql5.com/ru/docs/integration/python_metatrader5/mt5symbolinfotick_py

Yes, we should try it. Partly, I already got a blessing from fxsaber)

 
Aleksey Nikolayev:

Yes, we should try it. Partly, I already got fxsaber's blessing)

I already did it, when I tested the MO library, suggestedby mytarmailS. It worked like clockwork, but the TS was doomed from the beginning )

the script runs, in the loop it checks the arrival of a new hour (or another timeframe) and the logic is turning, everything is very simple
 
Maxim Dmitrievsky:

I already did, when I was testing the MO library suggestedby mytarmailS. Everything worked like clockwork, only the TC itself was initially doomed.)

the script is running, in the loop it checks the arrival of a new hour (or another timeframe) and the logic is turning, everything is very simple

What about testing it? In a homebrew python tester?

 
Aleksey Nikolayev:

What about testing? In a homemade python tester?

Yes, you can take it from my articles. Well, the tester is the easiest one. I can send you samples of bot on python.
Reason: