jaguar1637 liked this 4775 days ago
I open this discussion about the Better Clone Support Vector Machine (SVM) EA because I would like to hear is someone has any expericence with the code on this site:
Any suggestions are welcome.
I am opening the discussion because I made those indicators work and I input them into the Fractal Scalper EA.
Whe I am trying to read the available source code, yes it really looks like svm code. I insert the dll compiled file into libraries, the header file into include folder and the indicator into indicators folder.
(And I do not forget to set the allow dll tam, otherwise it will not work).
Then after that I called the SVM indicator into my EA. And as there are no parameters to optimize I just passed some test. Yes, it looks like this is working.
However I do not know if the tests are in sample or out of sample. I do not know how it is trained and for how long. I see that as input it is used the simple Moving average but with different periods.
OK here it is what I get.
![]() |
От Поле за пускане |
![]() |
От Поле за пускане |
I am amazed the one of the most sophisticated things as EA has gathered so little interest, when it was published.
The problem is that it is too complex and you can't trade complex things unless you are confident in the algorythm.
I am stll trying to understand the logic behind the indicators and where and what I can touch.
That is why I am not publishing my code yet. I tell you what it is, the idea is simple to use the SVM indicators calling them into an EA. I used the fractal scalper EA for calling them.
There are no optimization parameters whatsoever.
I do not know how the training happens and if the dll contains a legitimate svm code at all.
Well
Which indicator did you take inside the indicators folders ?
Which settings of this indicators ?
How do you call the SVM from your EA ?
It crashed my metatrader yesterday, I was panicked thinking that hard drive is gone.
I realized that I had to delete the indicators to make it work.
The integration is very simple just calling the icustom function.
I will upload it but it is purely experimental.