ForumSections

Machine learning in trading: theory, models, practice and algo-trading - page 2703

New comment
mytarmailS #:
Isn't it possible to test it? Do you have to trade for half a year?

Well, so as not to be tempted to fit the history.

Aleksey Vyazmikin #:

Well, so you're not tempted to make it fit the story.

I have no motive or temptation, it's a self-delusion.
mytarmailS #:
I have no motive or temptation to cheat myself.

It's not plausible - you want to"compete", not just compare results - hence the conditions I suggested.

I am interested in comparing results, so I can test it on history. Can you think of a function to activate the model, or should I suggest it?

Aleksey Vyazmikin #:

Not plausible - you want to"compete", not just compare results - hence the terms I suggested.

I am interested in comparing results, so I can test it on history. Can you think of a model activation function, or should I suggest it?

Model activation function?
I don't feel like competing anymore).
mytarmailS #:
Model activation function?
I don't feel like competing anymore )

Suggest your own variant - impulsiveness is unproductive.

Aleksey Vyazmikin #:

Suggest your own version - impulsiveness is unproductive.

For example...
Take a pattern and teach the model to predict whether the pattern will work or not...
Make a dataset for it. There won't be a lot of data, and that's good.
That's what you're doing, as I recall. I think you're doing it right.

I suggest we take a pattern of a price bounce from the mashka, or whatever you like....

mytarmailS #:
For example.
Take a pattern and teach the model to predict whether the pattern will work or not...
Make a dataset for it. There won't be a lot of data, and that's good..
That's what you're doing, as I recall. I think you're doing it right.

I suggest you take the price bounce pattern, or any pattern you like....

So this is in my understanding the function of model activation - a strict rule on the occurrence of which the model starts and produces a forecast.

Then maybe you should take the strategy from my article?

//+-----------------------------------------------------------------+
//| Возвращает сигнал на покупку или продажу - базовая стратегия
//+-----------------------------------------------------------------+
bool Signal()
{
//Сбрасываем Флаг блокировки открытия позиций
   SellPrIMA=false;  //Открывать отложенный ордер на продажу
   BuyPrIMA=false;   //Открывать отложенный ордер на покупку
   SellNow=false;    //Открывать ордер с рынка на продажу
   BuyNow=false;     //Открывать ордер с рынка на покупку
   bool Signal=false;//результат работы функции
   int BarN=0;       //Число баров без касания МА
   if(iOpen(Symbol(),Signal_MA_TF,0)>MA_Signal(0) && iLow(Symbol(),Signal_MA_TF,1)>MA_Signal(1))
   {
      for(int i=2; i<100; i++)
      {
         if(iLow(Symbol(),Signal_MA_TF,i)>MA_Signal(i))break;//На этом цикле уже был отработан сигнал
         if(iClose(Symbol(),Signal_MA_TF,i+1)<MA_Signal(i+1) && iClose(Symbol(),Signal_MA_TF,i)>=MA_Signal(i))
         {
            for(int x=i+1; x<100; x++)
            {
               if(iLow(Symbol(),Signal_MA_TF,x)>MA_Signal(x))break;//На этом цикле уже был отработан сигнал
               if(iHigh(Symbol(),Signal_MA_TF,x)<MA_Signal(x))
               {
                  BarN=x;
                  BuyNow=true;
                  break;
               }
            }
         }
      }
   }
   if(iOpen(Symbol(),Signal_MA_TF,0)<MA_Signal(0) && iHigh(Symbol(),Signal_MA_TF,1)<MA_Signal(1))
   {
      for(int i=2; i<100; i++)
      {
         if(iHigh(Symbol(),Signal_MA_TF,i)<MA_Signal(i))break;//На этом цикле уже был отработан сигнал
         if(iClose(Symbol(),Signal_MA_TF,i+1)>MA_Signal(i+1) && iClose(Symbol(),Signal_MA_TF,i)<=MA_Signal(i))
         {
            for(int x=i+1; x<100; x++)
            {
               if(iHigh(Symbol(),Signal_MA_TF,x)<MA_Signal(x))break;//На этом цикле уже был отработан сигнал
               if(iLow(Symbol(),Signal_MA_TF,x)>MA_Signal(x))
               {
                  BarN=x;
                  SellNow=true;
                  break;
               }
            }
         }
      }
   }
   if(BuyNow==true || SellNow==true)Signal=true;
   return Signal;
}
//+------------------------------------------------------------------+
//|Получим значение буфера индикатора handle_MA_Signal               |
//+------------------------------------------------------------------+
double MA_Signal(int index)
{
   double MA[1];
   ResetLastError();
   if(CopyBuffer(handle_MA_Signal,0,index,1,MA)<0)
   {
      PrintFormat("Failed to copy data from the handle_MA_Signal indicator, error code %d",GetLastError());
      return(0.0);
   }
//return NormalizeDouble(MA[0],Digits());
   return MA[0];
}


And in general, you can use that Expert Advisor - well, you will improve it for yourself, the binding to R, but there will be similarity in the results of the decision points.

Aleksey Vyazmikin #:

So this is in my understanding the function of model activation - a strict rule on the occurrence of which the model starts and produces a forecast.

Then maybe we should take the strategy from my article?


And in general, you can use that Expert Advisor - well, you can refine it for yourself, linkage to R, but there will be similarity in the results of the decision points.

Can you just post a CSV file for training (with answers) and for the test (without answers)?
elibrarius #:
Can't you just post a CSV file for the training (with answers) and for the test (without answers)?

That way everyone has different predictors - that's the problem.

If it's a question of who is better at using what they have - that's a different matter.

It is necessary to start with the same understanding of terms and goals in setting tasks. At this stage of communication try not to get syncrosia, i.e. allergy to the interlocutor, and only if this stage is passed without syncrosia can we hope for results.

Zy. it is a difficult task to assemble a team of people you know, and it is even more difficult to assemble a team of people you don't know))).

1...269626972698269927002701270227032704270527062707270827092710...3749
New comment