Machine learning in trading: theory, models, practice and algo-trading - page 2705

New comment
 
Aleksey Vyazmikin #:

I didn't understand from the articles how it is reproduced.

In the future we look at a random number of bars in a given range (from 5 to 20 for example), mark buy or sell on the current one, depending on what was in the future, go through the whole history like this
 
Maxim Dmitrievsky #:
In the future, we look at a random number of bars, mark buy or sell on the current one, depending on what was in the future, we go through the whole history in this way

I understand with the markup. But how do you apply the model in real trading, on each bar or with a cycle that randomly appeared on the markup, if with a cycle, how do you determine the beginning of the cycle, because the inclusion of the Expert Advisor can be at any time.

 
Aleksey Vyazmikin #:

I understand about the markup. But how do you apply the model, on each bar or with a cycle, which is randomly obtained on the markup, if with a cycle, how do you determine the beginning of the cycle, because the inclusion of the Expert Advisor can be at any time.

The model trades at any time, it has been trained
There are no cycles anywhere
 
Maxim Dmitrievsky #:
The model is anytime and trades, it's trained
There are no cycles anywhere

So the markup was on every bar so with delta from the future as a random number of bars?

 
Aleksey Vyazmikin #:

So the markup was on each bar means with delta from the future as a random number of bars?

Yes, but I don't think that's the only correct way to do it.
You can label it however you want, but the message was different.
 
Maxim Dmitrievsky #:
Yes

Does closing on a rollover signal mean?

 
Aleksey Vyazmikin #:

And closing on a rollover signal means?

Well
 
Maxim Dmitrievsky #:
Well.

That makes sense, thanks for the clarification.

 
Aleksey Vyazmikin #:
I'm reading here, I see that everyone understands their own conversations...
One saw a command, the other saw something else...

Karoch. The case should be like this: there is a common raw dataset with initial data and target...
In tsv or txt, so that anyone can take and do something from any language...
Predictors should be done by everyone, that's why it's raw...

I will not touch mcl, either you explain the initial rule (fn. Active. According to you) or I do it on mashka as I said before....

The purpose of this is to compare methods of generating traits, not a team, not to make money.
 
mytarmailS #:
I'm reading here, I see that everyone understands their own conversations...
One saw the team, the other saw something else.

Karoch. The case should be like this: there is a common raw dataset with initial data and target...
In csv or txt, so that anyone can take it and do something from any language...
Predictors to make for themselves everyone should himself, that's why it's raw...

I will not touch mcl, either you explain the initial rule (fn. Active. According to you) or I do on mashka as I said before....

The purpose of this is to compare methods of generating traits, not a team, not to make money.

You can't compare feature generation methods because I haven't created a system in code yet. What can be compared is your system with my set of predictors and my system/methodology for selecting them.

Anyone can get data from the historical interval of the MQL server - you want a continuous history. But the final sample to be trained on will be an order of magnitude smaller sample strings, but with additional predictors.

The Expert Advisor that I propose to use will save the open predictors and at the end of the csv file there will be columns with the financial result and target - you can take information on the time of triggering the "initial rule"/activation function from there, so there is no need to reproduce the algorithm in R.

I suggest the time interval - from 2010 to 2020 - training, the rest of the time for testing the results outside the training.

When you create your predictors, you can save the result in csv - and I will do so. Further or you can merge columns and study on different ranges or just separately - it is necessary for comparison of correctness of synchronisation.

I can send purely markup, if you don't want to get into it at all.

1...269826992700270127022703270427052706270727082709271027112712...3498
New comment
Reason: